xref: /linux/fs/f2fs/dir.c (revision 6d61a53dd6f55405ebcaea6ee38d1ab5a8856c2c)
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 <linux/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 BIT(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 #if IS_ENABLED(CONFIG_UNICODE)
46 /* If @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
47 int f2fs_init_casefolded_name(const struct inode *dir,
48 			      struct f2fs_filename *fname)
49 {
50 	struct super_block *sb = dir->i_sb;
51 	unsigned char *buf;
52 	int len;
53 
54 	if (IS_CASEFOLDED(dir) &&
55 	    !is_dot_dotdot(fname->usr_fname->name, fname->usr_fname->len)) {
56 		buf = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
57 					    GFP_NOFS, false, F2FS_SB(sb));
58 		if (!buf)
59 			return -ENOMEM;
60 
61 		len = utf8_casefold(sb->s_encoding, fname->usr_fname,
62 				    buf, F2FS_NAME_LEN);
63 		if (len <= 0) {
64 			kmem_cache_free(f2fs_cf_name_slab, buf);
65 			if (sb_has_strict_encoding(sb))
66 				return -EINVAL;
67 			/* fall back to treating name as opaque byte sequence */
68 			return 0;
69 		}
70 		fname->cf_name.name = buf;
71 		fname->cf_name.len = len;
72 	}
73 
74 	return 0;
75 }
76 
77 void f2fs_free_casefolded_name(struct f2fs_filename *fname)
78 {
79 	unsigned char *buf = (unsigned char *)fname->cf_name.name;
80 
81 	if (buf) {
82 		kmem_cache_free(f2fs_cf_name_slab, buf);
83 		fname->cf_name.name = NULL;
84 	}
85 }
86 #endif /* CONFIG_UNICODE */
87 
88 static int __f2fs_setup_filename(const struct inode *dir,
89 				 const struct fscrypt_name *crypt_name,
90 				 struct f2fs_filename *fname)
91 {
92 	int err;
93 
94 	memset(fname, 0, sizeof(*fname));
95 
96 	fname->usr_fname = crypt_name->usr_fname;
97 	fname->disk_name = crypt_name->disk_name;
98 #ifdef CONFIG_FS_ENCRYPTION
99 	fname->crypto_buf = crypt_name->crypto_buf;
100 #endif
101 	if (crypt_name->is_nokey_name) {
102 		/* hash was decoded from the no-key name */
103 		fname->hash = cpu_to_le32(crypt_name->hash);
104 	} else {
105 		err = f2fs_init_casefolded_name(dir, fname);
106 		if (err) {
107 			f2fs_free_filename(fname);
108 			return err;
109 		}
110 		f2fs_hash_filename(dir, fname);
111 	}
112 	return 0;
113 }
114 
115 /*
116  * Prepare to search for @iname in @dir.  This is similar to
117  * fscrypt_setup_filename(), but this also handles computing the casefolded name
118  * and the f2fs dirhash if needed, then packing all the information about this
119  * filename up into a 'struct f2fs_filename'.
120  */
121 int f2fs_setup_filename(struct inode *dir, const struct qstr *iname,
122 			int lookup, struct f2fs_filename *fname)
123 {
124 	struct fscrypt_name crypt_name;
125 	int err;
126 
127 	err = fscrypt_setup_filename(dir, iname, lookup, &crypt_name);
128 	if (err)
129 		return err;
130 
131 	return __f2fs_setup_filename(dir, &crypt_name, fname);
132 }
133 
134 /*
135  * Prepare to look up @dentry in @dir.  This is similar to
136  * fscrypt_prepare_lookup(), 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_prepare_lookup(struct inode *dir, struct dentry *dentry,
141 			struct f2fs_filename *fname)
142 {
143 	struct fscrypt_name crypt_name;
144 	int err;
145 
146 	err = fscrypt_prepare_lookup(dir, dentry, &crypt_name);
147 	if (err)
148 		return err;
149 
150 	return __f2fs_setup_filename(dir, &crypt_name, fname);
151 }
152 
153 void f2fs_free_filename(struct f2fs_filename *fname)
154 {
155 #ifdef CONFIG_FS_ENCRYPTION
156 	kfree(fname->crypto_buf.name);
157 	fname->crypto_buf.name = NULL;
158 #endif
159 	f2fs_free_casefolded_name(fname);
160 }
161 
162 static unsigned long dir_block_index(unsigned int level,
163 				int dir_level, unsigned int idx)
164 {
165 	unsigned long i;
166 	unsigned long bidx = 0;
167 
168 	for (i = 0; i < level; i++)
169 		bidx += mul_u32_u32(dir_buckets(i, dir_level),
170 				    bucket_blocks(i));
171 	bidx += idx * bucket_blocks(level);
172 	return bidx;
173 }
174 
175 static struct f2fs_dir_entry *find_in_block(struct inode *dir,
176 				struct page *dentry_page,
177 				const struct f2fs_filename *fname,
178 				int *max_slots,
179 				bool use_hash)
180 {
181 	struct f2fs_dentry_block *dentry_blk;
182 	struct f2fs_dentry_ptr d;
183 
184 	dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
185 
186 	make_dentry_ptr_block(dir, &d, dentry_blk);
187 	return f2fs_find_target_dentry(&d, fname, max_slots, use_hash);
188 }
189 
190 static inline int f2fs_match_name(const struct inode *dir,
191 				   const struct f2fs_filename *fname,
192 				   const u8 *de_name, u32 de_name_len)
193 {
194 	struct fscrypt_name f;
195 
196 #if IS_ENABLED(CONFIG_UNICODE)
197 	if (fname->cf_name.name)
198 		return generic_ci_match(dir, fname->usr_fname,
199 					&fname->cf_name,
200 					de_name, de_name_len);
201 
202 #endif
203 	f.usr_fname = fname->usr_fname;
204 	f.disk_name = fname->disk_name;
205 #ifdef CONFIG_FS_ENCRYPTION
206 	f.crypto_buf = fname->crypto_buf;
207 #endif
208 	return fscrypt_match_name(&f, de_name, de_name_len);
209 }
210 
211 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
212 			const struct f2fs_filename *fname, int *max_slots,
213 			bool use_hash)
214 {
215 	struct f2fs_dir_entry *de;
216 	unsigned long bit_pos = 0;
217 	int max_len = 0;
218 	int res = 0;
219 
220 	if (max_slots)
221 		*max_slots = 0;
222 	while (bit_pos < d->max) {
223 		if (!test_bit_le(bit_pos, d->bitmap)) {
224 			bit_pos++;
225 			max_len++;
226 			continue;
227 		}
228 
229 		de = &d->dentry[bit_pos];
230 
231 		if (unlikely(!de->name_len)) {
232 			bit_pos++;
233 			continue;
234 		}
235 
236 		if (!use_hash || de->hash_code == fname->hash) {
237 			res = f2fs_match_name(d->inode, fname,
238 					      d->filename[bit_pos],
239 					      le16_to_cpu(de->name_len));
240 			if (res < 0)
241 				return ERR_PTR(res);
242 			if (res)
243 				goto found;
244 		}
245 
246 		if (max_slots && max_len > *max_slots)
247 			*max_slots = max_len;
248 		max_len = 0;
249 
250 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
251 	}
252 
253 	de = NULL;
254 found:
255 	if (max_slots && max_len > *max_slots)
256 		*max_slots = max_len;
257 	return de;
258 }
259 
260 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
261 					unsigned int level,
262 					const struct f2fs_filename *fname,
263 					struct page **res_page,
264 					bool use_hash)
265 {
266 	int s = GET_DENTRY_SLOTS(fname->disk_name.len);
267 	unsigned int nbucket, nblock;
268 	unsigned int bidx, end_block, bucket_no;
269 	struct page *dentry_page;
270 	struct f2fs_dir_entry *de = NULL;
271 	pgoff_t next_pgofs;
272 	bool room = false;
273 	int max_slots;
274 
275 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
276 	nblock = bucket_blocks(level);
277 
278 	bucket_no = use_hash ? le32_to_cpu(fname->hash) % nbucket : 0;
279 
280 start_find_bucket:
281 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
282 			       bucket_no);
283 	end_block = bidx + nblock;
284 
285 	while (bidx < end_block) {
286 		/* no need to allocate new dentry pages to all the indices */
287 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
288 		if (IS_ERR(dentry_page)) {
289 			if (PTR_ERR(dentry_page) == -ENOENT) {
290 				room = true;
291 				bidx = next_pgofs;
292 				continue;
293 			} else {
294 				*res_page = dentry_page;
295 				break;
296 			}
297 		}
298 
299 		de = find_in_block(dir, dentry_page, fname, &max_slots, use_hash);
300 		if (IS_ERR(de)) {
301 			*res_page = ERR_CAST(de);
302 			de = NULL;
303 			break;
304 		} else if (de) {
305 			*res_page = dentry_page;
306 			break;
307 		}
308 
309 		if (max_slots >= s)
310 			room = true;
311 		f2fs_put_page(dentry_page, 0);
312 
313 		bidx++;
314 	}
315 
316 	if (de)
317 		return de;
318 
319 	if (likely(use_hash)) {
320 		if (room && F2FS_I(dir)->chash != fname->hash) {
321 			F2FS_I(dir)->chash = fname->hash;
322 			F2FS_I(dir)->clevel = level;
323 		}
324 	} else if (++bucket_no < nbucket) {
325 		goto start_find_bucket;
326 	}
327 	return NULL;
328 }
329 
330 struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
331 					 const struct f2fs_filename *fname,
332 					 struct page **res_page)
333 {
334 	unsigned long npages = dir_blocks(dir);
335 	struct f2fs_dir_entry *de = NULL;
336 	unsigned int max_depth;
337 	unsigned int level;
338 	bool use_hash = true;
339 
340 	*res_page = NULL;
341 
342 #if IS_ENABLED(CONFIG_UNICODE)
343 start_find_entry:
344 #endif
345 	if (f2fs_has_inline_dentry(dir)) {
346 		de = f2fs_find_in_inline_dir(dir, fname, res_page, use_hash);
347 		goto out;
348 	}
349 
350 	if (npages == 0)
351 		goto out;
352 
353 	max_depth = F2FS_I(dir)->i_current_depth;
354 	if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
355 		f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
356 			  dir->i_ino, max_depth);
357 		max_depth = MAX_DIR_HASH_DEPTH;
358 		f2fs_i_depth_write(dir, max_depth);
359 	}
360 
361 	for (level = 0; level < max_depth; level++) {
362 		de = find_in_level(dir, level, fname, res_page, use_hash);
363 		if (de || IS_ERR(*res_page))
364 			break;
365 	}
366 
367 out:
368 #if IS_ENABLED(CONFIG_UNICODE)
369 	if (IS_CASEFOLDED(dir) && !de && use_hash) {
370 		use_hash = false;
371 		goto start_find_entry;
372 	}
373 #endif
374 	/* This is to increase the speed of f2fs_create */
375 	if (!de)
376 		F2FS_I(dir)->task = current;
377 	return de;
378 }
379 
380 /*
381  * Find an entry in the specified directory with the wanted name.
382  * It returns the page where the entry was found (as a parameter - res_page),
383  * and the entry itself. Page is returned mapped and unlocked.
384  * Entry is guaranteed to be valid.
385  */
386 struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
387 			const struct qstr *child, struct page **res_page)
388 {
389 	struct f2fs_dir_entry *de = NULL;
390 	struct f2fs_filename fname;
391 	int err;
392 
393 	err = f2fs_setup_filename(dir, child, 1, &fname);
394 	if (err) {
395 		if (err == -ENOENT)
396 			*res_page = NULL;
397 		else
398 			*res_page = ERR_PTR(err);
399 		return NULL;
400 	}
401 
402 	de = __f2fs_find_entry(dir, &fname, res_page);
403 
404 	f2fs_free_filename(&fname);
405 	return de;
406 }
407 
408 struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
409 {
410 	return f2fs_find_entry(dir, &dotdot_name, p);
411 }
412 
413 ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
414 							struct page **page)
415 {
416 	ino_t res = 0;
417 	struct f2fs_dir_entry *de;
418 
419 	de = f2fs_find_entry(dir, qstr, page);
420 	if (de) {
421 		res = le32_to_cpu(de->ino);
422 		f2fs_put_page(*page, 0);
423 	}
424 
425 	return res;
426 }
427 
428 void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
429 		struct page *page, struct inode *inode)
430 {
431 	enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
432 
433 	lock_page(page);
434 	f2fs_wait_on_page_writeback(page, type, true, true);
435 	de->ino = cpu_to_le32(inode->i_ino);
436 	de->file_type = fs_umode_to_ftype(inode->i_mode);
437 	set_page_dirty(page);
438 
439 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
440 	f2fs_mark_inode_dirty_sync(dir, false);
441 	f2fs_put_page(page, 1);
442 }
443 
444 static void init_dent_inode(struct inode *dir, struct inode *inode,
445 			    const struct f2fs_filename *fname,
446 			    struct page *ipage)
447 {
448 	struct f2fs_inode *ri;
449 
450 	if (!fname) /* tmpfile case? */
451 		return;
452 
453 	f2fs_wait_on_page_writeback(ipage, NODE, true, true);
454 
455 	/* copy name info. to this inode page */
456 	ri = F2FS_INODE(ipage);
457 	ri->i_namelen = cpu_to_le32(fname->disk_name.len);
458 	memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
459 	if (IS_ENCRYPTED(dir)) {
460 		file_set_enc_name(inode);
461 		/*
462 		 * Roll-forward recovery doesn't have encryption keys available,
463 		 * so it can't compute the dirhash for encrypted+casefolded
464 		 * filenames.  Append it to i_name if possible.  Else, disable
465 		 * roll-forward recovery of the dentry (i.e., make fsync'ing the
466 		 * file force a checkpoint) by setting LOST_PINO.
467 		 */
468 		if (IS_CASEFOLDED(dir)) {
469 			if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
470 			    F2FS_NAME_LEN)
471 				put_unaligned(fname->hash, (f2fs_hash_t *)
472 					&ri->i_name[fname->disk_name.len]);
473 			else
474 				file_lost_pino(inode);
475 		}
476 	}
477 	set_page_dirty(ipage);
478 }
479 
480 void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
481 					struct f2fs_dentry_ptr *d)
482 {
483 	struct fscrypt_str dot = FSTR_INIT(".", 1);
484 	struct fscrypt_str dotdot = FSTR_INIT("..", 2);
485 
486 	/* update dirent of "." */
487 	f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
488 
489 	/* update dirent of ".." */
490 	f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
491 }
492 
493 static int make_empty_dir(struct inode *inode,
494 		struct inode *parent, struct page *page)
495 {
496 	struct page *dentry_page;
497 	struct f2fs_dentry_block *dentry_blk;
498 	struct f2fs_dentry_ptr d;
499 
500 	if (f2fs_has_inline_dentry(inode))
501 		return f2fs_make_empty_inline_dir(inode, parent, page);
502 
503 	dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
504 	if (IS_ERR(dentry_page))
505 		return PTR_ERR(dentry_page);
506 
507 	dentry_blk = page_address(dentry_page);
508 
509 	make_dentry_ptr_block(NULL, &d, dentry_blk);
510 	f2fs_do_make_empty_dir(inode, parent, &d);
511 
512 	set_page_dirty(dentry_page);
513 	f2fs_put_page(dentry_page, 1);
514 	return 0;
515 }
516 
517 struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
518 			const struct f2fs_filename *fname, struct page *dpage)
519 {
520 	struct page *page;
521 	int err;
522 
523 	if (is_inode_flag_set(inode, FI_NEW_INODE)) {
524 		page = f2fs_new_inode_page(inode);
525 		if (IS_ERR(page))
526 			return page;
527 
528 		if (S_ISDIR(inode->i_mode)) {
529 			/* in order to handle error case */
530 			get_page(page);
531 			err = make_empty_dir(inode, dir, page);
532 			if (err) {
533 				lock_page(page);
534 				goto put_error;
535 			}
536 			put_page(page);
537 		}
538 
539 		err = f2fs_init_acl(inode, dir, page, dpage);
540 		if (err)
541 			goto put_error;
542 
543 		err = f2fs_init_security(inode, dir,
544 					 fname ? fname->usr_fname : NULL, page);
545 		if (err)
546 			goto put_error;
547 
548 		if (IS_ENCRYPTED(inode)) {
549 			err = fscrypt_set_context(inode, page);
550 			if (err)
551 				goto put_error;
552 		}
553 	} else {
554 		page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
555 		if (IS_ERR(page))
556 			return page;
557 	}
558 
559 	init_dent_inode(dir, inode, fname, page);
560 
561 	/*
562 	 * This file should be checkpointed during fsync.
563 	 * We lost i_pino from now on.
564 	 */
565 	if (is_inode_flag_set(inode, FI_INC_LINK)) {
566 		if (!S_ISDIR(inode->i_mode))
567 			file_lost_pino(inode);
568 		/*
569 		 * If link the tmpfile to alias through linkat path,
570 		 * we should remove this inode from orphan list.
571 		 */
572 		if (inode->i_nlink == 0)
573 			f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
574 		f2fs_i_links_write(inode, true);
575 	}
576 	return page;
577 
578 put_error:
579 	clear_nlink(inode);
580 	f2fs_update_inode(inode, page);
581 	f2fs_put_page(page, 1);
582 	return ERR_PTR(err);
583 }
584 
585 void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
586 						unsigned int current_depth)
587 {
588 	if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
589 		if (S_ISDIR(inode->i_mode))
590 			f2fs_i_links_write(dir, true);
591 		clear_inode_flag(inode, FI_NEW_INODE);
592 	}
593 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
594 	f2fs_mark_inode_dirty_sync(dir, false);
595 
596 	if (F2FS_I(dir)->i_current_depth != current_depth)
597 		f2fs_i_depth_write(dir, current_depth);
598 
599 	if (inode && is_inode_flag_set(inode, FI_INC_LINK))
600 		clear_inode_flag(inode, FI_INC_LINK);
601 }
602 
603 int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
604 {
605 	int bit_start = 0;
606 	int zero_start, zero_end;
607 next:
608 	zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
609 	if (zero_start >= max_slots)
610 		return max_slots;
611 
612 	zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
613 	if (zero_end - zero_start >= slots)
614 		return zero_start;
615 
616 	bit_start = zero_end + 1;
617 
618 	if (zero_end + 1 >= max_slots)
619 		return max_slots;
620 	goto next;
621 }
622 
623 bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
624 			  const struct f2fs_filename *fname)
625 {
626 	struct f2fs_dentry_ptr d;
627 	unsigned int bit_pos;
628 	int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
629 
630 	make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
631 
632 	bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
633 
634 	return bit_pos < d.max;
635 }
636 
637 void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
638 			const struct fscrypt_str *name, f2fs_hash_t name_hash,
639 			unsigned int bit_pos)
640 {
641 	struct f2fs_dir_entry *de;
642 	int slots = GET_DENTRY_SLOTS(name->len);
643 	int i;
644 
645 	de = &d->dentry[bit_pos];
646 	de->hash_code = name_hash;
647 	de->name_len = cpu_to_le16(name->len);
648 	memcpy(d->filename[bit_pos], name->name, name->len);
649 	de->ino = cpu_to_le32(ino);
650 	de->file_type = fs_umode_to_ftype(mode);
651 	for (i = 0; i < slots; i++) {
652 		__set_bit_le(bit_pos + i, (void *)d->bitmap);
653 		/* avoid wrong garbage data for readdir */
654 		if (i)
655 			(de + i)->name_len = 0;
656 	}
657 }
658 
659 int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
660 			   struct inode *inode, nid_t ino, umode_t mode)
661 {
662 	unsigned int bit_pos;
663 	unsigned int level;
664 	unsigned int current_depth;
665 	unsigned long bidx, block;
666 	unsigned int nbucket, nblock;
667 	struct page *dentry_page = NULL;
668 	struct f2fs_dentry_block *dentry_blk = NULL;
669 	struct f2fs_dentry_ptr d;
670 	struct page *page = NULL;
671 	int slots, err = 0;
672 
673 	level = 0;
674 	slots = GET_DENTRY_SLOTS(fname->disk_name.len);
675 
676 	current_depth = F2FS_I(dir)->i_current_depth;
677 	if (F2FS_I(dir)->chash == fname->hash) {
678 		level = F2FS_I(dir)->clevel;
679 		F2FS_I(dir)->chash = 0;
680 	}
681 
682 start:
683 	if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH))
684 		return -ENOSPC;
685 
686 	if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
687 		return -ENOSPC;
688 
689 	/* Increase the depth, if required */
690 	if (level == current_depth)
691 		++current_depth;
692 
693 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
694 	nblock = bucket_blocks(level);
695 
696 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
697 				(le32_to_cpu(fname->hash) % nbucket));
698 
699 	for (block = bidx; block <= (bidx + nblock - 1); block++) {
700 		dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
701 		if (IS_ERR(dentry_page))
702 			return PTR_ERR(dentry_page);
703 
704 		dentry_blk = page_address(dentry_page);
705 		bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
706 						slots, NR_DENTRY_IN_BLOCK);
707 		if (bit_pos < NR_DENTRY_IN_BLOCK)
708 			goto add_dentry;
709 
710 		f2fs_put_page(dentry_page, 1);
711 	}
712 
713 	/* Move to next level to find the empty slot for new dentry */
714 	++level;
715 	goto start;
716 add_dentry:
717 	f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
718 
719 	if (inode) {
720 		f2fs_down_write(&F2FS_I(inode)->i_sem);
721 		page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
722 		if (IS_ERR(page)) {
723 			err = PTR_ERR(page);
724 			goto fail;
725 		}
726 	}
727 
728 	make_dentry_ptr_block(NULL, &d, dentry_blk);
729 	f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
730 			   bit_pos);
731 
732 	set_page_dirty(dentry_page);
733 
734 	if (inode) {
735 		f2fs_i_pino_write(inode, dir->i_ino);
736 
737 		/* synchronize inode page's data from inode cache */
738 		if (is_inode_flag_set(inode, FI_NEW_INODE))
739 			f2fs_update_inode(inode, page);
740 
741 		f2fs_put_page(page, 1);
742 	}
743 
744 	f2fs_update_parent_metadata(dir, inode, current_depth);
745 fail:
746 	if (inode)
747 		f2fs_up_write(&F2FS_I(inode)->i_sem);
748 
749 	f2fs_put_page(dentry_page, 1);
750 
751 	return err;
752 }
753 
754 int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
755 		    struct inode *inode, nid_t ino, umode_t mode)
756 {
757 	int err = -EAGAIN;
758 
759 	if (f2fs_has_inline_dentry(dir)) {
760 		/*
761 		 * Should get i_xattr_sem to keep the lock order:
762 		 * i_xattr_sem -> inode_page lock used by f2fs_setxattr.
763 		 */
764 		f2fs_down_read(&F2FS_I(dir)->i_xattr_sem);
765 		err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
766 		f2fs_up_read(&F2FS_I(dir)->i_xattr_sem);
767 	}
768 	if (err == -EAGAIN)
769 		err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
770 
771 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
772 	return err;
773 }
774 
775 /*
776  * Caller should grab and release a rwsem by calling f2fs_lock_op() and
777  * f2fs_unlock_op().
778  */
779 int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
780 				struct inode *inode, nid_t ino, umode_t mode)
781 {
782 	struct f2fs_filename fname;
783 	struct page *page = NULL;
784 	struct f2fs_dir_entry *de = NULL;
785 	int err;
786 
787 	err = f2fs_setup_filename(dir, name, 0, &fname);
788 	if (err)
789 		return err;
790 
791 	/*
792 	 * An immature stackable filesystem shows a race condition between lookup
793 	 * and create. If we have same task when doing lookup and create, it's
794 	 * definitely fine as expected by VFS normally. Otherwise, let's just
795 	 * verify on-disk dentry one more time, which guarantees filesystem
796 	 * consistency more.
797 	 */
798 	if (current != F2FS_I(dir)->task) {
799 		de = __f2fs_find_entry(dir, &fname, &page);
800 		F2FS_I(dir)->task = NULL;
801 	}
802 	if (de) {
803 		f2fs_put_page(page, 0);
804 		err = -EEXIST;
805 	} else if (IS_ERR(page)) {
806 		err = PTR_ERR(page);
807 	} else {
808 		err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
809 	}
810 	f2fs_free_filename(&fname);
811 	return err;
812 }
813 
814 int f2fs_do_tmpfile(struct inode *inode, struct inode *dir,
815 					struct f2fs_filename *fname)
816 {
817 	struct page *page;
818 	int err = 0;
819 
820 	f2fs_down_write(&F2FS_I(inode)->i_sem);
821 	page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
822 	if (IS_ERR(page)) {
823 		err = PTR_ERR(page);
824 		goto fail;
825 	}
826 	f2fs_put_page(page, 1);
827 
828 	clear_inode_flag(inode, FI_NEW_INODE);
829 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
830 fail:
831 	f2fs_up_write(&F2FS_I(inode)->i_sem);
832 	return err;
833 }
834 
835 void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
836 {
837 	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
838 
839 	f2fs_down_write(&F2FS_I(inode)->i_sem);
840 
841 	if (S_ISDIR(inode->i_mode))
842 		f2fs_i_links_write(dir, false);
843 	inode_set_ctime_current(inode);
844 
845 	f2fs_i_links_write(inode, false);
846 	if (S_ISDIR(inode->i_mode)) {
847 		f2fs_i_links_write(inode, false);
848 		f2fs_i_size_write(inode, 0);
849 	}
850 	f2fs_up_write(&F2FS_I(inode)->i_sem);
851 
852 	if (inode->i_nlink == 0)
853 		f2fs_add_orphan_inode(inode);
854 	else
855 		f2fs_release_orphan_inode(sbi);
856 }
857 
858 /*
859  * It only removes the dentry from the dentry page, corresponding name
860  * entry in name page does not need to be touched during deletion.
861  */
862 void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
863 					struct inode *dir, struct inode *inode)
864 {
865 	struct	f2fs_dentry_block *dentry_blk;
866 	unsigned int bit_pos;
867 	int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
868 	pgoff_t index = page_folio(page)->index;
869 	int i;
870 
871 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
872 
873 	if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
874 		f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
875 
876 	if (f2fs_has_inline_dentry(dir))
877 		return f2fs_delete_inline_entry(dentry, page, dir, inode);
878 
879 	lock_page(page);
880 	f2fs_wait_on_page_writeback(page, DATA, true, true);
881 
882 	dentry_blk = page_address(page);
883 	bit_pos = dentry - dentry_blk->dentry;
884 	for (i = 0; i < slots; i++)
885 		__clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
886 
887 	/* Let's check and deallocate this dentry page */
888 	bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
889 			NR_DENTRY_IN_BLOCK,
890 			0);
891 	set_page_dirty(page);
892 
893 	if (bit_pos == NR_DENTRY_IN_BLOCK &&
894 		!f2fs_truncate_hole(dir, index, index + 1)) {
895 		f2fs_clear_page_cache_dirty_tag(page_folio(page));
896 		clear_page_dirty_for_io(page);
897 		ClearPageUptodate(page);
898 		clear_page_private_all(page);
899 
900 		inode_dec_dirty_pages(dir);
901 		f2fs_remove_dirty_inode(dir);
902 	}
903 	f2fs_put_page(page, 1);
904 
905 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
906 	f2fs_mark_inode_dirty_sync(dir, false);
907 
908 	if (inode)
909 		f2fs_drop_nlink(dir, inode);
910 }
911 
912 bool f2fs_empty_dir(struct inode *dir)
913 {
914 	unsigned long bidx = 0;
915 	struct page *dentry_page;
916 	unsigned int bit_pos;
917 	struct f2fs_dentry_block *dentry_blk;
918 	unsigned long nblock = dir_blocks(dir);
919 
920 	if (f2fs_has_inline_dentry(dir))
921 		return f2fs_empty_inline_dir(dir);
922 
923 	while (bidx < nblock) {
924 		pgoff_t next_pgofs;
925 
926 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
927 		if (IS_ERR(dentry_page)) {
928 			if (PTR_ERR(dentry_page) == -ENOENT) {
929 				bidx = next_pgofs;
930 				continue;
931 			} else {
932 				return false;
933 			}
934 		}
935 
936 		dentry_blk = page_address(dentry_page);
937 		if (bidx == 0)
938 			bit_pos = 2;
939 		else
940 			bit_pos = 0;
941 		bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
942 						NR_DENTRY_IN_BLOCK,
943 						bit_pos);
944 
945 		f2fs_put_page(dentry_page, 0);
946 
947 		if (bit_pos < NR_DENTRY_IN_BLOCK)
948 			return false;
949 
950 		bidx++;
951 	}
952 	return true;
953 }
954 
955 int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
956 			unsigned int start_pos, struct fscrypt_str *fstr)
957 {
958 	unsigned char d_type = DT_UNKNOWN;
959 	unsigned int bit_pos;
960 	struct f2fs_dir_entry *de = NULL;
961 	struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
962 	struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
963 	struct blk_plug plug;
964 	bool readdir_ra = sbi->readdir_ra;
965 	bool found_valid_dirent = false;
966 	int err = 0;
967 
968 	bit_pos = ((unsigned long)ctx->pos % d->max);
969 
970 	if (readdir_ra)
971 		blk_start_plug(&plug);
972 
973 	while (bit_pos < d->max) {
974 		bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
975 		if (bit_pos >= d->max)
976 			break;
977 
978 		de = &d->dentry[bit_pos];
979 		if (de->name_len == 0) {
980 			if (found_valid_dirent || !bit_pos) {
981 				f2fs_warn_ratelimited(sbi,
982 					"invalid namelen(0), ino:%u, run fsck to fix.",
983 					le32_to_cpu(de->ino));
984 				set_sbi_flag(sbi, SBI_NEED_FSCK);
985 			}
986 			bit_pos++;
987 			ctx->pos = start_pos + bit_pos;
988 			continue;
989 		}
990 
991 		d_type = fs_ftype_to_dtype(de->file_type);
992 
993 		de_name.name = d->filename[bit_pos];
994 		de_name.len = le16_to_cpu(de->name_len);
995 
996 		/* check memory boundary before moving forward */
997 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
998 		if (unlikely(bit_pos > d->max ||
999 				le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
1000 			f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
1001 				  __func__, le16_to_cpu(de->name_len));
1002 			set_sbi_flag(sbi, SBI_NEED_FSCK);
1003 			err = -EFSCORRUPTED;
1004 			f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
1005 			goto out;
1006 		}
1007 
1008 		if (IS_ENCRYPTED(d->inode)) {
1009 			int save_len = fstr->len;
1010 
1011 			err = fscrypt_fname_disk_to_usr(d->inode,
1012 						(u32)le32_to_cpu(de->hash_code),
1013 						0, &de_name, fstr);
1014 			if (err)
1015 				goto out;
1016 
1017 			de_name = *fstr;
1018 			fstr->len = save_len;
1019 		}
1020 
1021 		if (!dir_emit(ctx, de_name.name, de_name.len,
1022 					le32_to_cpu(de->ino), d_type)) {
1023 			err = 1;
1024 			goto out;
1025 		}
1026 
1027 		if (readdir_ra)
1028 			f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1029 
1030 		ctx->pos = start_pos + bit_pos;
1031 		found_valid_dirent = true;
1032 	}
1033 out:
1034 	if (readdir_ra)
1035 		blk_finish_plug(&plug);
1036 	return err;
1037 }
1038 
1039 static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1040 {
1041 	struct inode *inode = file_inode(file);
1042 	unsigned long npages = dir_blocks(inode);
1043 	struct f2fs_dentry_block *dentry_blk = NULL;
1044 	struct page *dentry_page = NULL;
1045 	struct file_ra_state *ra = &file->f_ra;
1046 	loff_t start_pos = ctx->pos;
1047 	unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1048 	struct f2fs_dentry_ptr d;
1049 	struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1050 	int err = 0;
1051 
1052 	if (IS_ENCRYPTED(inode)) {
1053 		err = fscrypt_prepare_readdir(inode);
1054 		if (err)
1055 			goto out;
1056 
1057 		err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1058 		if (err < 0)
1059 			goto out;
1060 	}
1061 
1062 	if (f2fs_has_inline_dentry(inode)) {
1063 		err = f2fs_read_inline_dir(file, ctx, &fstr);
1064 		goto out_free;
1065 	}
1066 
1067 	for (; n < npages; ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1068 		pgoff_t next_pgofs;
1069 
1070 		/* allow readdir() to be interrupted */
1071 		if (fatal_signal_pending(current)) {
1072 			err = -ERESTARTSYS;
1073 			goto out_free;
1074 		}
1075 		cond_resched();
1076 
1077 		/* readahead for multi pages of dir */
1078 		if (npages - n > 1 && !ra_has_index(ra, n))
1079 			page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1080 				min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1081 
1082 		dentry_page = f2fs_find_data_page(inode, n, &next_pgofs);
1083 		if (IS_ERR(dentry_page)) {
1084 			err = PTR_ERR(dentry_page);
1085 			if (err == -ENOENT) {
1086 				err = 0;
1087 				n = next_pgofs;
1088 				continue;
1089 			} else {
1090 				goto out_free;
1091 			}
1092 		}
1093 
1094 		dentry_blk = page_address(dentry_page);
1095 
1096 		make_dentry_ptr_block(inode, &d, dentry_blk);
1097 
1098 		err = f2fs_fill_dentries(ctx, &d,
1099 				n * NR_DENTRY_IN_BLOCK, &fstr);
1100 		if (err) {
1101 			f2fs_put_page(dentry_page, 0);
1102 			break;
1103 		}
1104 
1105 		f2fs_put_page(dentry_page, 0);
1106 
1107 		n++;
1108 	}
1109 out_free:
1110 	fscrypt_fname_free_buffer(&fstr);
1111 out:
1112 	trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1113 	return err < 0 ? err : 0;
1114 }
1115 
1116 const struct file_operations f2fs_dir_operations = {
1117 	.llseek		= generic_file_llseek,
1118 	.read		= generic_read_dir,
1119 	.iterate_shared	= f2fs_readdir,
1120 	.fsync		= f2fs_sync_file,
1121 	.unlocked_ioctl	= f2fs_ioctl,
1122 #ifdef CONFIG_COMPAT
1123 	.compat_ioctl   = f2fs_compat_ioctl,
1124 #endif
1125 };
1126