xref: /openbmc/linux/fs/f2fs/dir.c (revision d88e2a2b)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * fs/f2fs/dir.c
4  *
5  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6  *             http://www.samsung.com/
7  */
8 #include <asm/unaligned.h>
9 #include <linux/fs.h>
10 #include <linux/f2fs_fs.h>
11 #include <linux/sched/signal.h>
12 #include <linux/unicode.h>
13 #include "f2fs.h"
14 #include "node.h"
15 #include "acl.h"
16 #include "xattr.h"
17 #include <trace/events/f2fs.h>
18 
19 #if IS_ENABLED(CONFIG_UNICODE)
20 extern struct kmem_cache *f2fs_cf_name_slab;
21 #endif
22 
23 static unsigned long dir_blocks(struct inode *inode)
24 {
25 	return ((unsigned long long) (i_size_read(inode) + PAGE_SIZE - 1))
26 							>> PAGE_SHIFT;
27 }
28 
29 static unsigned int dir_buckets(unsigned int level, int dir_level)
30 {
31 	if (level + dir_level < MAX_DIR_HASH_DEPTH / 2)
32 		return 1 << (level + dir_level);
33 	else
34 		return MAX_DIR_BUCKETS;
35 }
36 
37 static unsigned int bucket_blocks(unsigned int level)
38 {
39 	if (level < MAX_DIR_HASH_DEPTH / 2)
40 		return 2;
41 	else
42 		return 4;
43 }
44 
45 static unsigned char f2fs_filetype_table[F2FS_FT_MAX] = {
46 	[F2FS_FT_UNKNOWN]	= DT_UNKNOWN,
47 	[F2FS_FT_REG_FILE]	= DT_REG,
48 	[F2FS_FT_DIR]		= DT_DIR,
49 	[F2FS_FT_CHRDEV]	= DT_CHR,
50 	[F2FS_FT_BLKDEV]	= DT_BLK,
51 	[F2FS_FT_FIFO]		= DT_FIFO,
52 	[F2FS_FT_SOCK]		= DT_SOCK,
53 	[F2FS_FT_SYMLINK]	= DT_LNK,
54 };
55 
56 static unsigned char f2fs_type_by_mode[S_IFMT >> S_SHIFT] = {
57 	[S_IFREG >> S_SHIFT]	= F2FS_FT_REG_FILE,
58 	[S_IFDIR >> S_SHIFT]	= F2FS_FT_DIR,
59 	[S_IFCHR >> S_SHIFT]	= F2FS_FT_CHRDEV,
60 	[S_IFBLK >> S_SHIFT]	= F2FS_FT_BLKDEV,
61 	[S_IFIFO >> S_SHIFT]	= F2FS_FT_FIFO,
62 	[S_IFSOCK >> S_SHIFT]	= F2FS_FT_SOCK,
63 	[S_IFLNK >> S_SHIFT]	= F2FS_FT_SYMLINK,
64 };
65 
66 static void set_de_type(struct f2fs_dir_entry *de, umode_t mode)
67 {
68 	de->file_type = f2fs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
69 }
70 
71 unsigned char f2fs_get_de_type(struct f2fs_dir_entry *de)
72 {
73 	if (de->file_type < F2FS_FT_MAX)
74 		return f2fs_filetype_table[de->file_type];
75 	return DT_UNKNOWN;
76 }
77 
78 /* If @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
79 int f2fs_init_casefolded_name(const struct inode *dir,
80 			      struct f2fs_filename *fname)
81 {
82 #if IS_ENABLED(CONFIG_UNICODE)
83 	struct super_block *sb = dir->i_sb;
84 
85 	if (IS_CASEFOLDED(dir) &&
86 	    !is_dot_dotdot(fname->usr_fname->name, fname->usr_fname->len)) {
87 		fname->cf_name.name = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
88 					GFP_NOFS, false, F2FS_SB(sb));
89 		if (!fname->cf_name.name)
90 			return -ENOMEM;
91 		fname->cf_name.len = utf8_casefold(sb->s_encoding,
92 						   fname->usr_fname,
93 						   fname->cf_name.name,
94 						   F2FS_NAME_LEN);
95 		if ((int)fname->cf_name.len <= 0) {
96 			kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
97 			fname->cf_name.name = NULL;
98 			if (sb_has_strict_encoding(sb))
99 				return -EINVAL;
100 			/* fall back to treating name as opaque byte sequence */
101 		}
102 	}
103 #endif
104 	return 0;
105 }
106 
107 static int __f2fs_setup_filename(const struct inode *dir,
108 				 const struct fscrypt_name *crypt_name,
109 				 struct f2fs_filename *fname)
110 {
111 	int err;
112 
113 	memset(fname, 0, sizeof(*fname));
114 
115 	fname->usr_fname = crypt_name->usr_fname;
116 	fname->disk_name = crypt_name->disk_name;
117 #ifdef CONFIG_FS_ENCRYPTION
118 	fname->crypto_buf = crypt_name->crypto_buf;
119 #endif
120 	if (crypt_name->is_nokey_name) {
121 		/* hash was decoded from the no-key name */
122 		fname->hash = cpu_to_le32(crypt_name->hash);
123 	} else {
124 		err = f2fs_init_casefolded_name(dir, fname);
125 		if (err) {
126 			f2fs_free_filename(fname);
127 			return err;
128 		}
129 		f2fs_hash_filename(dir, fname);
130 	}
131 	return 0;
132 }
133 
134 /*
135  * Prepare to search for @iname in @dir.  This is similar to
136  * fscrypt_setup_filename(), but this also handles computing the casefolded name
137  * and the f2fs dirhash if needed, then packing all the information about this
138  * filename up into a 'struct f2fs_filename'.
139  */
140 int f2fs_setup_filename(struct inode *dir, const struct qstr *iname,
141 			int lookup, struct f2fs_filename *fname)
142 {
143 	struct fscrypt_name crypt_name;
144 	int err;
145 
146 	err = fscrypt_setup_filename(dir, iname, lookup, &crypt_name);
147 	if (err)
148 		return err;
149 
150 	return __f2fs_setup_filename(dir, &crypt_name, fname);
151 }
152 
153 /*
154  * Prepare to look up @dentry in @dir.  This is similar to
155  * fscrypt_prepare_lookup(), but this also handles computing the casefolded name
156  * and the f2fs dirhash if needed, then packing all the information about this
157  * filename up into a 'struct f2fs_filename'.
158  */
159 int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
160 			struct f2fs_filename *fname)
161 {
162 	struct fscrypt_name crypt_name;
163 	int err;
164 
165 	err = fscrypt_prepare_lookup(dir, dentry, &crypt_name);
166 	if (err)
167 		return err;
168 
169 	return __f2fs_setup_filename(dir, &crypt_name, fname);
170 }
171 
172 void f2fs_free_filename(struct f2fs_filename *fname)
173 {
174 #ifdef CONFIG_FS_ENCRYPTION
175 	kfree(fname->crypto_buf.name);
176 	fname->crypto_buf.name = NULL;
177 #endif
178 #if IS_ENABLED(CONFIG_UNICODE)
179 	if (fname->cf_name.name) {
180 		kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
181 		fname->cf_name.name = NULL;
182 	}
183 #endif
184 }
185 
186 static unsigned long dir_block_index(unsigned int level,
187 				int dir_level, unsigned int idx)
188 {
189 	unsigned long i;
190 	unsigned long bidx = 0;
191 
192 	for (i = 0; i < level; i++)
193 		bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
194 	bidx += idx * bucket_blocks(level);
195 	return bidx;
196 }
197 
198 static struct f2fs_dir_entry *find_in_block(struct inode *dir,
199 				struct page *dentry_page,
200 				const struct f2fs_filename *fname,
201 				int *max_slots)
202 {
203 	struct f2fs_dentry_block *dentry_blk;
204 	struct f2fs_dentry_ptr d;
205 
206 	dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
207 
208 	make_dentry_ptr_block(dir, &d, dentry_blk);
209 	return f2fs_find_target_dentry(&d, fname, max_slots);
210 }
211 
212 #if IS_ENABLED(CONFIG_UNICODE)
213 /*
214  * Test whether a case-insensitive directory entry matches the filename
215  * being searched for.
216  *
217  * Returns 1 for a match, 0 for no match, and -errno on an error.
218  */
219 static int f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
220 			       const u8 *de_name, u32 de_name_len)
221 {
222 	const struct super_block *sb = dir->i_sb;
223 	const struct unicode_map *um = sb->s_encoding;
224 	struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
225 	struct qstr entry = QSTR_INIT(de_name, de_name_len);
226 	int res;
227 
228 	if (IS_ENCRYPTED(dir)) {
229 		const struct fscrypt_str encrypted_name =
230 			FSTR_INIT((u8 *)de_name, de_name_len);
231 
232 		if (WARN_ON_ONCE(!fscrypt_has_encryption_key(dir)))
233 			return -EINVAL;
234 
235 		decrypted_name.name = kmalloc(de_name_len, GFP_KERNEL);
236 		if (!decrypted_name.name)
237 			return -ENOMEM;
238 		res = fscrypt_fname_disk_to_usr(dir, 0, 0, &encrypted_name,
239 						&decrypted_name);
240 		if (res < 0)
241 			goto out;
242 		entry.name = decrypted_name.name;
243 		entry.len = decrypted_name.len;
244 	}
245 
246 	res = utf8_strncasecmp_folded(um, name, &entry);
247 	/*
248 	 * In strict mode, ignore invalid names.  In non-strict mode,
249 	 * fall back to treating them as opaque byte sequences.
250 	 */
251 	if (res < 0 && !sb_has_strict_encoding(sb)) {
252 		res = name->len == entry.len &&
253 				memcmp(name->name, entry.name, name->len) == 0;
254 	} else {
255 		/* utf8_strncasecmp_folded returns 0 on match */
256 		res = (res == 0);
257 	}
258 out:
259 	kfree(decrypted_name.name);
260 	return res;
261 }
262 #endif /* CONFIG_UNICODE */
263 
264 static inline int f2fs_match_name(const struct inode *dir,
265 				   const struct f2fs_filename *fname,
266 				   const u8 *de_name, u32 de_name_len)
267 {
268 	struct fscrypt_name f;
269 
270 #if IS_ENABLED(CONFIG_UNICODE)
271 	if (fname->cf_name.name) {
272 		struct qstr cf = FSTR_TO_QSTR(&fname->cf_name);
273 
274 		return f2fs_match_ci_name(dir, &cf, de_name, de_name_len);
275 	}
276 #endif
277 	f.usr_fname = fname->usr_fname;
278 	f.disk_name = fname->disk_name;
279 #ifdef CONFIG_FS_ENCRYPTION
280 	f.crypto_buf = fname->crypto_buf;
281 #endif
282 	return fscrypt_match_name(&f, de_name, de_name_len);
283 }
284 
285 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
286 			const struct f2fs_filename *fname, int *max_slots)
287 {
288 	struct f2fs_dir_entry *de;
289 	unsigned long bit_pos = 0;
290 	int max_len = 0;
291 	int res = 0;
292 
293 	if (max_slots)
294 		*max_slots = 0;
295 	while (bit_pos < d->max) {
296 		if (!test_bit_le(bit_pos, d->bitmap)) {
297 			bit_pos++;
298 			max_len++;
299 			continue;
300 		}
301 
302 		de = &d->dentry[bit_pos];
303 
304 		if (unlikely(!de->name_len)) {
305 			bit_pos++;
306 			continue;
307 		}
308 
309 		if (de->hash_code == fname->hash) {
310 			res = f2fs_match_name(d->inode, fname,
311 					      d->filename[bit_pos],
312 					      le16_to_cpu(de->name_len));
313 			if (res < 0)
314 				return ERR_PTR(res);
315 			if (res)
316 				goto found;
317 		}
318 
319 		if (max_slots && max_len > *max_slots)
320 			*max_slots = max_len;
321 		max_len = 0;
322 
323 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
324 	}
325 
326 	de = NULL;
327 found:
328 	if (max_slots && max_len > *max_slots)
329 		*max_slots = max_len;
330 	return de;
331 }
332 
333 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
334 					unsigned int level,
335 					const struct f2fs_filename *fname,
336 					struct page **res_page)
337 {
338 	int s = GET_DENTRY_SLOTS(fname->disk_name.len);
339 	unsigned int nbucket, nblock;
340 	unsigned int bidx, end_block;
341 	struct page *dentry_page;
342 	struct f2fs_dir_entry *de = NULL;
343 	pgoff_t next_pgofs;
344 	bool room = false;
345 	int max_slots;
346 
347 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
348 	nblock = bucket_blocks(level);
349 
350 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
351 			       le32_to_cpu(fname->hash) % nbucket);
352 	end_block = bidx + nblock;
353 
354 	while (bidx < end_block) {
355 		/* no need to allocate new dentry pages to all the indices */
356 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
357 		if (IS_ERR(dentry_page)) {
358 			if (PTR_ERR(dentry_page) == -ENOENT) {
359 				room = true;
360 				bidx = next_pgofs;
361 				continue;
362 			} else {
363 				*res_page = dentry_page;
364 				break;
365 			}
366 		}
367 
368 		de = find_in_block(dir, dentry_page, fname, &max_slots);
369 		if (IS_ERR(de)) {
370 			*res_page = ERR_CAST(de);
371 			de = NULL;
372 			break;
373 		} else if (de) {
374 			*res_page = dentry_page;
375 			break;
376 		}
377 
378 		if (max_slots >= s)
379 			room = true;
380 		f2fs_put_page(dentry_page, 0);
381 
382 		bidx++;
383 	}
384 
385 	if (!de && room && F2FS_I(dir)->chash != fname->hash) {
386 		F2FS_I(dir)->chash = fname->hash;
387 		F2FS_I(dir)->clevel = level;
388 	}
389 
390 	return de;
391 }
392 
393 struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
394 					 const struct f2fs_filename *fname,
395 					 struct page **res_page)
396 {
397 	unsigned long npages = dir_blocks(dir);
398 	struct f2fs_dir_entry *de = NULL;
399 	unsigned int max_depth;
400 	unsigned int level;
401 
402 	*res_page = NULL;
403 
404 	if (f2fs_has_inline_dentry(dir)) {
405 		de = f2fs_find_in_inline_dir(dir, fname, res_page);
406 		goto out;
407 	}
408 
409 	if (npages == 0)
410 		goto out;
411 
412 	max_depth = F2FS_I(dir)->i_current_depth;
413 	if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
414 		f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
415 			  dir->i_ino, max_depth);
416 		max_depth = MAX_DIR_HASH_DEPTH;
417 		f2fs_i_depth_write(dir, max_depth);
418 	}
419 
420 	for (level = 0; level < max_depth; level++) {
421 		de = find_in_level(dir, level, fname, res_page);
422 		if (de || IS_ERR(*res_page))
423 			break;
424 	}
425 out:
426 	/* This is to increase the speed of f2fs_create */
427 	if (!de)
428 		F2FS_I(dir)->task = current;
429 	return de;
430 }
431 
432 /*
433  * Find an entry in the specified directory with the wanted name.
434  * It returns the page where the entry was found (as a parameter - res_page),
435  * and the entry itself. Page is returned mapped and unlocked.
436  * Entry is guaranteed to be valid.
437  */
438 struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
439 			const struct qstr *child, struct page **res_page)
440 {
441 	struct f2fs_dir_entry *de = NULL;
442 	struct f2fs_filename fname;
443 	int err;
444 
445 	err = f2fs_setup_filename(dir, child, 1, &fname);
446 	if (err) {
447 		if (err == -ENOENT)
448 			*res_page = NULL;
449 		else
450 			*res_page = ERR_PTR(err);
451 		return NULL;
452 	}
453 
454 	de = __f2fs_find_entry(dir, &fname, res_page);
455 
456 	f2fs_free_filename(&fname);
457 	return de;
458 }
459 
460 struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
461 {
462 	return f2fs_find_entry(dir, &dotdot_name, p);
463 }
464 
465 ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
466 							struct page **page)
467 {
468 	ino_t res = 0;
469 	struct f2fs_dir_entry *de;
470 
471 	de = f2fs_find_entry(dir, qstr, page);
472 	if (de) {
473 		res = le32_to_cpu(de->ino);
474 		f2fs_put_page(*page, 0);
475 	}
476 
477 	return res;
478 }
479 
480 void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
481 		struct page *page, struct inode *inode)
482 {
483 	enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
484 
485 	lock_page(page);
486 	f2fs_wait_on_page_writeback(page, type, true, true);
487 	de->ino = cpu_to_le32(inode->i_ino);
488 	set_de_type(de, inode->i_mode);
489 	set_page_dirty(page);
490 
491 	dir->i_mtime = dir->i_ctime = current_time(dir);
492 	f2fs_mark_inode_dirty_sync(dir, false);
493 	f2fs_put_page(page, 1);
494 }
495 
496 static void init_dent_inode(struct inode *dir, struct inode *inode,
497 			    const struct f2fs_filename *fname,
498 			    struct page *ipage)
499 {
500 	struct f2fs_inode *ri;
501 
502 	if (!fname) /* tmpfile case? */
503 		return;
504 
505 	f2fs_wait_on_page_writeback(ipage, NODE, true, true);
506 
507 	/* copy name info. to this inode page */
508 	ri = F2FS_INODE(ipage);
509 	ri->i_namelen = cpu_to_le32(fname->disk_name.len);
510 	memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
511 	if (IS_ENCRYPTED(dir)) {
512 		file_set_enc_name(inode);
513 		/*
514 		 * Roll-forward recovery doesn't have encryption keys available,
515 		 * so it can't compute the dirhash for encrypted+casefolded
516 		 * filenames.  Append it to i_name if possible.  Else, disable
517 		 * roll-forward recovery of the dentry (i.e., make fsync'ing the
518 		 * file force a checkpoint) by setting LOST_PINO.
519 		 */
520 		if (IS_CASEFOLDED(dir)) {
521 			if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
522 			    F2FS_NAME_LEN)
523 				put_unaligned(fname->hash, (f2fs_hash_t *)
524 					&ri->i_name[fname->disk_name.len]);
525 			else
526 				file_lost_pino(inode);
527 		}
528 	}
529 	set_page_dirty(ipage);
530 }
531 
532 void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
533 					struct f2fs_dentry_ptr *d)
534 {
535 	struct fscrypt_str dot = FSTR_INIT(".", 1);
536 	struct fscrypt_str dotdot = FSTR_INIT("..", 2);
537 
538 	/* update dirent of "." */
539 	f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
540 
541 	/* update dirent of ".." */
542 	f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
543 }
544 
545 static int make_empty_dir(struct inode *inode,
546 		struct inode *parent, struct page *page)
547 {
548 	struct page *dentry_page;
549 	struct f2fs_dentry_block *dentry_blk;
550 	struct f2fs_dentry_ptr d;
551 
552 	if (f2fs_has_inline_dentry(inode))
553 		return f2fs_make_empty_inline_dir(inode, parent, page);
554 
555 	dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
556 	if (IS_ERR(dentry_page))
557 		return PTR_ERR(dentry_page);
558 
559 	dentry_blk = page_address(dentry_page);
560 
561 	make_dentry_ptr_block(NULL, &d, dentry_blk);
562 	f2fs_do_make_empty_dir(inode, parent, &d);
563 
564 	set_page_dirty(dentry_page);
565 	f2fs_put_page(dentry_page, 1);
566 	return 0;
567 }
568 
569 struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
570 			const struct f2fs_filename *fname, struct page *dpage)
571 {
572 	struct page *page;
573 	int err;
574 
575 	if (is_inode_flag_set(inode, FI_NEW_INODE)) {
576 		page = f2fs_new_inode_page(inode);
577 		if (IS_ERR(page))
578 			return page;
579 
580 		if (S_ISDIR(inode->i_mode)) {
581 			/* in order to handle error case */
582 			get_page(page);
583 			err = make_empty_dir(inode, dir, page);
584 			if (err) {
585 				lock_page(page);
586 				goto put_error;
587 			}
588 			put_page(page);
589 		}
590 
591 		err = f2fs_init_acl(inode, dir, page, dpage);
592 		if (err)
593 			goto put_error;
594 
595 		err = f2fs_init_security(inode, dir,
596 					 fname ? fname->usr_fname : NULL, page);
597 		if (err)
598 			goto put_error;
599 
600 		if (IS_ENCRYPTED(inode)) {
601 			err = fscrypt_set_context(inode, page);
602 			if (err)
603 				goto put_error;
604 		}
605 	} else {
606 		page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
607 		if (IS_ERR(page))
608 			return page;
609 	}
610 
611 	init_dent_inode(dir, inode, fname, page);
612 
613 	/*
614 	 * This file should be checkpointed during fsync.
615 	 * We lost i_pino from now on.
616 	 */
617 	if (is_inode_flag_set(inode, FI_INC_LINK)) {
618 		if (!S_ISDIR(inode->i_mode))
619 			file_lost_pino(inode);
620 		/*
621 		 * If link the tmpfile to alias through linkat path,
622 		 * we should remove this inode from orphan list.
623 		 */
624 		if (inode->i_nlink == 0)
625 			f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
626 		f2fs_i_links_write(inode, true);
627 	}
628 	return page;
629 
630 put_error:
631 	clear_nlink(inode);
632 	f2fs_update_inode(inode, page);
633 	f2fs_put_page(page, 1);
634 	return ERR_PTR(err);
635 }
636 
637 void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
638 						unsigned int current_depth)
639 {
640 	if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
641 		if (S_ISDIR(inode->i_mode))
642 			f2fs_i_links_write(dir, true);
643 		clear_inode_flag(inode, FI_NEW_INODE);
644 	}
645 	dir->i_mtime = dir->i_ctime = current_time(dir);
646 	f2fs_mark_inode_dirty_sync(dir, false);
647 
648 	if (F2FS_I(dir)->i_current_depth != current_depth)
649 		f2fs_i_depth_write(dir, current_depth);
650 
651 	if (inode && is_inode_flag_set(inode, FI_INC_LINK))
652 		clear_inode_flag(inode, FI_INC_LINK);
653 }
654 
655 int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
656 {
657 	int bit_start = 0;
658 	int zero_start, zero_end;
659 next:
660 	zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
661 	if (zero_start >= max_slots)
662 		return max_slots;
663 
664 	zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
665 	if (zero_end - zero_start >= slots)
666 		return zero_start;
667 
668 	bit_start = zero_end + 1;
669 
670 	if (zero_end + 1 >= max_slots)
671 		return max_slots;
672 	goto next;
673 }
674 
675 bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
676 			  const struct f2fs_filename *fname)
677 {
678 	struct f2fs_dentry_ptr d;
679 	unsigned int bit_pos;
680 	int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
681 
682 	make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
683 
684 	bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
685 
686 	return bit_pos < d.max;
687 }
688 
689 void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
690 			const struct fscrypt_str *name, f2fs_hash_t name_hash,
691 			unsigned int bit_pos)
692 {
693 	struct f2fs_dir_entry *de;
694 	int slots = GET_DENTRY_SLOTS(name->len);
695 	int i;
696 
697 	de = &d->dentry[bit_pos];
698 	de->hash_code = name_hash;
699 	de->name_len = cpu_to_le16(name->len);
700 	memcpy(d->filename[bit_pos], name->name, name->len);
701 	de->ino = cpu_to_le32(ino);
702 	set_de_type(de, mode);
703 	for (i = 0; i < slots; i++) {
704 		__set_bit_le(bit_pos + i, (void *)d->bitmap);
705 		/* avoid wrong garbage data for readdir */
706 		if (i)
707 			(de + i)->name_len = 0;
708 	}
709 }
710 
711 int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
712 			   struct inode *inode, nid_t ino, umode_t mode)
713 {
714 	unsigned int bit_pos;
715 	unsigned int level;
716 	unsigned int current_depth;
717 	unsigned long bidx, block;
718 	unsigned int nbucket, nblock;
719 	struct page *dentry_page = NULL;
720 	struct f2fs_dentry_block *dentry_blk = NULL;
721 	struct f2fs_dentry_ptr d;
722 	struct page *page = NULL;
723 	int slots, err = 0;
724 
725 	level = 0;
726 	slots = GET_DENTRY_SLOTS(fname->disk_name.len);
727 
728 	current_depth = F2FS_I(dir)->i_current_depth;
729 	if (F2FS_I(dir)->chash == fname->hash) {
730 		level = F2FS_I(dir)->clevel;
731 		F2FS_I(dir)->chash = 0;
732 	}
733 
734 start:
735 	if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH))
736 		return -ENOSPC;
737 
738 	if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
739 		return -ENOSPC;
740 
741 	/* Increase the depth, if required */
742 	if (level == current_depth)
743 		++current_depth;
744 
745 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
746 	nblock = bucket_blocks(level);
747 
748 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
749 				(le32_to_cpu(fname->hash) % nbucket));
750 
751 	for (block = bidx; block <= (bidx + nblock - 1); block++) {
752 		dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
753 		if (IS_ERR(dentry_page))
754 			return PTR_ERR(dentry_page);
755 
756 		dentry_blk = page_address(dentry_page);
757 		bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
758 						slots, NR_DENTRY_IN_BLOCK);
759 		if (bit_pos < NR_DENTRY_IN_BLOCK)
760 			goto add_dentry;
761 
762 		f2fs_put_page(dentry_page, 1);
763 	}
764 
765 	/* Move to next level to find the empty slot for new dentry */
766 	++level;
767 	goto start;
768 add_dentry:
769 	f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
770 
771 	if (inode) {
772 		f2fs_down_write(&F2FS_I(inode)->i_sem);
773 		page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
774 		if (IS_ERR(page)) {
775 			err = PTR_ERR(page);
776 			goto fail;
777 		}
778 	}
779 
780 	make_dentry_ptr_block(NULL, &d, dentry_blk);
781 	f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
782 			   bit_pos);
783 
784 	set_page_dirty(dentry_page);
785 
786 	if (inode) {
787 		f2fs_i_pino_write(inode, dir->i_ino);
788 
789 		/* synchronize inode page's data from inode cache */
790 		if (is_inode_flag_set(inode, FI_NEW_INODE))
791 			f2fs_update_inode(inode, page);
792 
793 		f2fs_put_page(page, 1);
794 	}
795 
796 	f2fs_update_parent_metadata(dir, inode, current_depth);
797 fail:
798 	if (inode)
799 		f2fs_up_write(&F2FS_I(inode)->i_sem);
800 
801 	f2fs_put_page(dentry_page, 1);
802 
803 	return err;
804 }
805 
806 int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
807 		    struct inode *inode, nid_t ino, umode_t mode)
808 {
809 	int err = -EAGAIN;
810 
811 	if (f2fs_has_inline_dentry(dir))
812 		err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
813 	if (err == -EAGAIN)
814 		err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
815 
816 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
817 	return err;
818 }
819 
820 /*
821  * Caller should grab and release a rwsem by calling f2fs_lock_op() and
822  * f2fs_unlock_op().
823  */
824 int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
825 				struct inode *inode, nid_t ino, umode_t mode)
826 {
827 	struct f2fs_filename fname;
828 	struct page *page = NULL;
829 	struct f2fs_dir_entry *de = NULL;
830 	int err;
831 
832 	err = f2fs_setup_filename(dir, name, 0, &fname);
833 	if (err)
834 		return err;
835 
836 	/*
837 	 * An immature stackable filesystem shows a race condition between lookup
838 	 * and create. If we have same task when doing lookup and create, it's
839 	 * definitely fine as expected by VFS normally. Otherwise, let's just
840 	 * verify on-disk dentry one more time, which guarantees filesystem
841 	 * consistency more.
842 	 */
843 	if (current != F2FS_I(dir)->task) {
844 		de = __f2fs_find_entry(dir, &fname, &page);
845 		F2FS_I(dir)->task = NULL;
846 	}
847 	if (de) {
848 		f2fs_put_page(page, 0);
849 		err = -EEXIST;
850 	} else if (IS_ERR(page)) {
851 		err = PTR_ERR(page);
852 	} else {
853 		err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
854 	}
855 	f2fs_free_filename(&fname);
856 	return err;
857 }
858 
859 int f2fs_do_tmpfile(struct inode *inode, struct inode *dir)
860 {
861 	struct page *page;
862 	int err = 0;
863 
864 	f2fs_down_write(&F2FS_I(inode)->i_sem);
865 	page = f2fs_init_inode_metadata(inode, dir, NULL, NULL);
866 	if (IS_ERR(page)) {
867 		err = PTR_ERR(page);
868 		goto fail;
869 	}
870 	f2fs_put_page(page, 1);
871 
872 	clear_inode_flag(inode, FI_NEW_INODE);
873 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
874 fail:
875 	f2fs_up_write(&F2FS_I(inode)->i_sem);
876 	return err;
877 }
878 
879 void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
880 {
881 	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
882 
883 	f2fs_down_write(&F2FS_I(inode)->i_sem);
884 
885 	if (S_ISDIR(inode->i_mode))
886 		f2fs_i_links_write(dir, false);
887 	inode->i_ctime = current_time(inode);
888 
889 	f2fs_i_links_write(inode, false);
890 	if (S_ISDIR(inode->i_mode)) {
891 		f2fs_i_links_write(inode, false);
892 		f2fs_i_size_write(inode, 0);
893 	}
894 	f2fs_up_write(&F2FS_I(inode)->i_sem);
895 
896 	if (inode->i_nlink == 0)
897 		f2fs_add_orphan_inode(inode);
898 	else
899 		f2fs_release_orphan_inode(sbi);
900 }
901 
902 /*
903  * It only removes the dentry from the dentry page, corresponding name
904  * entry in name page does not need to be touched during deletion.
905  */
906 void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
907 					struct inode *dir, struct inode *inode)
908 {
909 	struct	f2fs_dentry_block *dentry_blk;
910 	unsigned int bit_pos;
911 	int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
912 	int i;
913 
914 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
915 
916 	if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
917 		f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
918 
919 	if (f2fs_has_inline_dentry(dir))
920 		return f2fs_delete_inline_entry(dentry, page, dir, inode);
921 
922 	lock_page(page);
923 	f2fs_wait_on_page_writeback(page, DATA, true, true);
924 
925 	dentry_blk = page_address(page);
926 	bit_pos = dentry - dentry_blk->dentry;
927 	for (i = 0; i < slots; i++)
928 		__clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
929 
930 	/* Let's check and deallocate this dentry page */
931 	bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
932 			NR_DENTRY_IN_BLOCK,
933 			0);
934 	set_page_dirty(page);
935 
936 	if (bit_pos == NR_DENTRY_IN_BLOCK &&
937 		!f2fs_truncate_hole(dir, page->index, page->index + 1)) {
938 		f2fs_clear_page_cache_dirty_tag(page);
939 		clear_page_dirty_for_io(page);
940 		ClearPageUptodate(page);
941 
942 		clear_page_private_gcing(page);
943 
944 		inode_dec_dirty_pages(dir);
945 		f2fs_remove_dirty_inode(dir);
946 
947 		detach_page_private(page);
948 		set_page_private(page, 0);
949 	}
950 	f2fs_put_page(page, 1);
951 
952 	dir->i_ctime = dir->i_mtime = current_time(dir);
953 	f2fs_mark_inode_dirty_sync(dir, false);
954 
955 	if (inode)
956 		f2fs_drop_nlink(dir, inode);
957 }
958 
959 bool f2fs_empty_dir(struct inode *dir)
960 {
961 	unsigned long bidx = 0;
962 	struct page *dentry_page;
963 	unsigned int bit_pos;
964 	struct f2fs_dentry_block *dentry_blk;
965 	unsigned long nblock = dir_blocks(dir);
966 
967 	if (f2fs_has_inline_dentry(dir))
968 		return f2fs_empty_inline_dir(dir);
969 
970 	while (bidx < nblock) {
971 		pgoff_t next_pgofs;
972 
973 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
974 		if (IS_ERR(dentry_page)) {
975 			if (PTR_ERR(dentry_page) == -ENOENT) {
976 				bidx = next_pgofs;
977 				continue;
978 			} else {
979 				return false;
980 			}
981 		}
982 
983 		dentry_blk = page_address(dentry_page);
984 		if (bidx == 0)
985 			bit_pos = 2;
986 		else
987 			bit_pos = 0;
988 		bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
989 						NR_DENTRY_IN_BLOCK,
990 						bit_pos);
991 
992 		f2fs_put_page(dentry_page, 0);
993 
994 		if (bit_pos < NR_DENTRY_IN_BLOCK)
995 			return false;
996 
997 		bidx++;
998 	}
999 	return true;
1000 }
1001 
1002 int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
1003 			unsigned int start_pos, struct fscrypt_str *fstr)
1004 {
1005 	unsigned char d_type = DT_UNKNOWN;
1006 	unsigned int bit_pos;
1007 	struct f2fs_dir_entry *de = NULL;
1008 	struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
1009 	struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
1010 	struct blk_plug plug;
1011 	bool readdir_ra = sbi->readdir_ra;
1012 	bool found_valid_dirent = false;
1013 	int err = 0;
1014 
1015 	bit_pos = ((unsigned long)ctx->pos % d->max);
1016 
1017 	if (readdir_ra)
1018 		blk_start_plug(&plug);
1019 
1020 	while (bit_pos < d->max) {
1021 		bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
1022 		if (bit_pos >= d->max)
1023 			break;
1024 
1025 		de = &d->dentry[bit_pos];
1026 		if (de->name_len == 0) {
1027 			if (found_valid_dirent || !bit_pos) {
1028 				printk_ratelimited(
1029 					"%sF2FS-fs (%s): invalid namelen(0), ino:%u, run fsck to fix.",
1030 					KERN_WARNING, sbi->sb->s_id,
1031 					le32_to_cpu(de->ino));
1032 				set_sbi_flag(sbi, SBI_NEED_FSCK);
1033 			}
1034 			bit_pos++;
1035 			ctx->pos = start_pos + bit_pos;
1036 			continue;
1037 		}
1038 
1039 		d_type = f2fs_get_de_type(de);
1040 
1041 		de_name.name = d->filename[bit_pos];
1042 		de_name.len = le16_to_cpu(de->name_len);
1043 
1044 		/* check memory boundary before moving forward */
1045 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
1046 		if (unlikely(bit_pos > d->max ||
1047 				le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
1048 			f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
1049 				  __func__, le16_to_cpu(de->name_len));
1050 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1051 			err = -EFSCORRUPTED;
1052 			f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
1053 			goto out;
1054 		}
1055 
1056 		if (IS_ENCRYPTED(d->inode)) {
1057 			int save_len = fstr->len;
1058 
1059 			err = fscrypt_fname_disk_to_usr(d->inode,
1060 						(u32)le32_to_cpu(de->hash_code),
1061 						0, &de_name, fstr);
1062 			if (err)
1063 				goto out;
1064 
1065 			de_name = *fstr;
1066 			fstr->len = save_len;
1067 		}
1068 
1069 		if (!dir_emit(ctx, de_name.name, de_name.len,
1070 					le32_to_cpu(de->ino), d_type)) {
1071 			err = 1;
1072 			goto out;
1073 		}
1074 
1075 		if (readdir_ra)
1076 			f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1077 
1078 		ctx->pos = start_pos + bit_pos;
1079 		found_valid_dirent = true;
1080 	}
1081 out:
1082 	if (readdir_ra)
1083 		blk_finish_plug(&plug);
1084 	return err;
1085 }
1086 
1087 static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1088 {
1089 	struct inode *inode = file_inode(file);
1090 	unsigned long npages = dir_blocks(inode);
1091 	struct f2fs_dentry_block *dentry_blk = NULL;
1092 	struct page *dentry_page = NULL;
1093 	struct file_ra_state *ra = &file->f_ra;
1094 	loff_t start_pos = ctx->pos;
1095 	unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1096 	struct f2fs_dentry_ptr d;
1097 	struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1098 	int err = 0;
1099 
1100 	if (IS_ENCRYPTED(inode)) {
1101 		err = fscrypt_prepare_readdir(inode);
1102 		if (err)
1103 			goto out;
1104 
1105 		err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1106 		if (err < 0)
1107 			goto out;
1108 	}
1109 
1110 	if (f2fs_has_inline_dentry(inode)) {
1111 		err = f2fs_read_inline_dir(file, ctx, &fstr);
1112 		goto out_free;
1113 	}
1114 
1115 	for (; n < npages; ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1116 		pgoff_t next_pgofs;
1117 
1118 		/* allow readdir() to be interrupted */
1119 		if (fatal_signal_pending(current)) {
1120 			err = -ERESTARTSYS;
1121 			goto out_free;
1122 		}
1123 		cond_resched();
1124 
1125 		/* readahead for multi pages of dir */
1126 		if (npages - n > 1 && !ra_has_index(ra, n))
1127 			page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1128 				min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1129 
1130 		dentry_page = f2fs_find_data_page(inode, n, &next_pgofs);
1131 		if (IS_ERR(dentry_page)) {
1132 			err = PTR_ERR(dentry_page);
1133 			if (err == -ENOENT) {
1134 				err = 0;
1135 				n = next_pgofs;
1136 				continue;
1137 			} else {
1138 				goto out_free;
1139 			}
1140 		}
1141 
1142 		dentry_blk = page_address(dentry_page);
1143 
1144 		make_dentry_ptr_block(inode, &d, dentry_blk);
1145 
1146 		err = f2fs_fill_dentries(ctx, &d,
1147 				n * NR_DENTRY_IN_BLOCK, &fstr);
1148 		if (err) {
1149 			f2fs_put_page(dentry_page, 0);
1150 			break;
1151 		}
1152 
1153 		f2fs_put_page(dentry_page, 0);
1154 
1155 		n++;
1156 	}
1157 out_free:
1158 	fscrypt_fname_free_buffer(&fstr);
1159 out:
1160 	trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1161 	return err < 0 ? err : 0;
1162 }
1163 
1164 const struct file_operations f2fs_dir_operations = {
1165 	.llseek		= generic_file_llseek,
1166 	.read		= generic_read_dir,
1167 	.iterate_shared	= f2fs_readdir,
1168 	.fsync		= f2fs_sync_file,
1169 	.unlocked_ioctl	= f2fs_ioctl,
1170 #ifdef CONFIG_COMPAT
1171 	.compat_ioctl   = f2fs_compat_ioctl,
1172 #endif
1173 };
1174