Loading...
1/*
2 * fs/f2fs/dir.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/fs.h>
12#include <linux/f2fs_fs.h>
13#include "f2fs.h"
14#include "node.h"
15#include "acl.h"
16#include "xattr.h"
17
18static unsigned long dir_blocks(struct inode *inode)
19{
20 return ((unsigned long long) (i_size_read(inode) + PAGE_CACHE_SIZE - 1))
21 >> PAGE_CACHE_SHIFT;
22}
23
24static unsigned int dir_buckets(unsigned int level, int dir_level)
25{
26 if (level < MAX_DIR_HASH_DEPTH / 2)
27 return 1 << (level + dir_level);
28 else
29 return 1 << ((MAX_DIR_HASH_DEPTH / 2 + dir_level) - 1);
30}
31
32static unsigned int bucket_blocks(unsigned int level)
33{
34 if (level < MAX_DIR_HASH_DEPTH / 2)
35 return 2;
36 else
37 return 4;
38}
39
40static unsigned char f2fs_filetype_table[F2FS_FT_MAX] = {
41 [F2FS_FT_UNKNOWN] = DT_UNKNOWN,
42 [F2FS_FT_REG_FILE] = DT_REG,
43 [F2FS_FT_DIR] = DT_DIR,
44 [F2FS_FT_CHRDEV] = DT_CHR,
45 [F2FS_FT_BLKDEV] = DT_BLK,
46 [F2FS_FT_FIFO] = DT_FIFO,
47 [F2FS_FT_SOCK] = DT_SOCK,
48 [F2FS_FT_SYMLINK] = DT_LNK,
49};
50
51#define S_SHIFT 12
52static unsigned char f2fs_type_by_mode[S_IFMT >> S_SHIFT] = {
53 [S_IFREG >> S_SHIFT] = F2FS_FT_REG_FILE,
54 [S_IFDIR >> S_SHIFT] = F2FS_FT_DIR,
55 [S_IFCHR >> S_SHIFT] = F2FS_FT_CHRDEV,
56 [S_IFBLK >> S_SHIFT] = F2FS_FT_BLKDEV,
57 [S_IFIFO >> S_SHIFT] = F2FS_FT_FIFO,
58 [S_IFSOCK >> S_SHIFT] = F2FS_FT_SOCK,
59 [S_IFLNK >> S_SHIFT] = F2FS_FT_SYMLINK,
60};
61
62static void set_de_type(struct f2fs_dir_entry *de, struct inode *inode)
63{
64 umode_t mode = inode->i_mode;
65 de->file_type = f2fs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
66}
67
68static unsigned long dir_block_index(unsigned int level,
69 int dir_level, unsigned int idx)
70{
71 unsigned long i;
72 unsigned long bidx = 0;
73
74 for (i = 0; i < level; i++)
75 bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
76 bidx += idx * bucket_blocks(level);
77 return bidx;
78}
79
80static bool early_match_name(const char *name, size_t namelen,
81 f2fs_hash_t namehash, struct f2fs_dir_entry *de)
82{
83 if (le16_to_cpu(de->name_len) != namelen)
84 return false;
85
86 if (de->hash_code != namehash)
87 return false;
88
89 return true;
90}
91
92static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
93 const char *name, size_t namelen, int *max_slots,
94 f2fs_hash_t namehash, struct page **res_page)
95{
96 struct f2fs_dir_entry *de;
97 unsigned long bit_pos = 0;
98 struct f2fs_dentry_block *dentry_blk = kmap(dentry_page);
99 const void *dentry_bits = &dentry_blk->dentry_bitmap;
100 int max_len = 0;
101
102 while (bit_pos < NR_DENTRY_IN_BLOCK) {
103 if (!test_bit_le(bit_pos, dentry_bits)) {
104 if (bit_pos == 0)
105 max_len = 1;
106 else if (!test_bit_le(bit_pos - 1, dentry_bits))
107 max_len++;
108 bit_pos++;
109 continue;
110 }
111 de = &dentry_blk->dentry[bit_pos];
112 if (early_match_name(name, namelen, namehash, de)) {
113 if (!memcmp(dentry_blk->filename[bit_pos],
114 name, namelen)) {
115 *res_page = dentry_page;
116 goto found;
117 }
118 }
119 if (max_len > *max_slots) {
120 *max_slots = max_len;
121 max_len = 0;
122 }
123 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
124 }
125
126 de = NULL;
127 kunmap(dentry_page);
128found:
129 if (max_len > *max_slots)
130 *max_slots = max_len;
131 return de;
132}
133
134static struct f2fs_dir_entry *find_in_level(struct inode *dir,
135 unsigned int level, const char *name, size_t namelen,
136 f2fs_hash_t namehash, struct page **res_page)
137{
138 int s = GET_DENTRY_SLOTS(namelen);
139 unsigned int nbucket, nblock;
140 unsigned int bidx, end_block;
141 struct page *dentry_page;
142 struct f2fs_dir_entry *de = NULL;
143 bool room = false;
144 int max_slots = 0;
145
146 f2fs_bug_on(level > MAX_DIR_HASH_DEPTH);
147
148 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
149 nblock = bucket_blocks(level);
150
151 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
152 le32_to_cpu(namehash) % nbucket);
153 end_block = bidx + nblock;
154
155 for (; bidx < end_block; bidx++) {
156 /* no need to allocate new dentry pages to all the indices */
157 dentry_page = find_data_page(dir, bidx, true);
158 if (IS_ERR(dentry_page)) {
159 room = true;
160 continue;
161 }
162
163 de = find_in_block(dentry_page, name, namelen,
164 &max_slots, namehash, res_page);
165 if (de)
166 break;
167
168 if (max_slots >= s)
169 room = true;
170 f2fs_put_page(dentry_page, 0);
171 }
172
173 if (!de && room && F2FS_I(dir)->chash != namehash) {
174 F2FS_I(dir)->chash = namehash;
175 F2FS_I(dir)->clevel = level;
176 }
177
178 return de;
179}
180
181/*
182 * Find an entry in the specified directory with the wanted name.
183 * It returns the page where the entry was found (as a parameter - res_page),
184 * and the entry itself. Page is returned mapped and unlocked.
185 * Entry is guaranteed to be valid.
186 */
187struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
188 struct qstr *child, struct page **res_page)
189{
190 const char *name = child->name;
191 size_t namelen = child->len;
192 unsigned long npages = dir_blocks(dir);
193 struct f2fs_dir_entry *de = NULL;
194 f2fs_hash_t name_hash;
195 unsigned int max_depth;
196 unsigned int level;
197
198 if (npages == 0)
199 return NULL;
200
201 *res_page = NULL;
202
203 name_hash = f2fs_dentry_hash(name, namelen);
204 max_depth = F2FS_I(dir)->i_current_depth;
205
206 for (level = 0; level < max_depth; level++) {
207 de = find_in_level(dir, level, name,
208 namelen, name_hash, res_page);
209 if (de)
210 break;
211 }
212 if (!de && F2FS_I(dir)->chash != name_hash) {
213 F2FS_I(dir)->chash = name_hash;
214 F2FS_I(dir)->clevel = level - 1;
215 }
216 return de;
217}
218
219struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
220{
221 struct page *page;
222 struct f2fs_dir_entry *de;
223 struct f2fs_dentry_block *dentry_blk;
224
225 page = get_lock_data_page(dir, 0);
226 if (IS_ERR(page))
227 return NULL;
228
229 dentry_blk = kmap(page);
230 de = &dentry_blk->dentry[1];
231 *p = page;
232 unlock_page(page);
233 return de;
234}
235
236ino_t f2fs_inode_by_name(struct inode *dir, struct qstr *qstr)
237{
238 ino_t res = 0;
239 struct f2fs_dir_entry *de;
240 struct page *page;
241
242 de = f2fs_find_entry(dir, qstr, &page);
243 if (de) {
244 res = le32_to_cpu(de->ino);
245 kunmap(page);
246 f2fs_put_page(page, 0);
247 }
248
249 return res;
250}
251
252void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
253 struct page *page, struct inode *inode)
254{
255 lock_page(page);
256 f2fs_wait_on_page_writeback(page, DATA);
257 de->ino = cpu_to_le32(inode->i_ino);
258 set_de_type(de, inode);
259 kunmap(page);
260 set_page_dirty(page);
261 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
262 mark_inode_dirty(dir);
263
264 f2fs_put_page(page, 1);
265}
266
267static void init_dent_inode(const struct qstr *name, struct page *ipage)
268{
269 struct f2fs_inode *ri;
270
271 /* copy name info. to this inode page */
272 ri = F2FS_INODE(ipage);
273 ri->i_namelen = cpu_to_le32(name->len);
274 memcpy(ri->i_name, name->name, name->len);
275 set_page_dirty(ipage);
276}
277
278int update_dent_inode(struct inode *inode, const struct qstr *name)
279{
280 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
281 struct page *page;
282
283 page = get_node_page(sbi, inode->i_ino);
284 if (IS_ERR(page))
285 return PTR_ERR(page);
286
287 init_dent_inode(name, page);
288 f2fs_put_page(page, 1);
289
290 return 0;
291}
292
293static int make_empty_dir(struct inode *inode,
294 struct inode *parent, struct page *page)
295{
296 struct page *dentry_page;
297 struct f2fs_dentry_block *dentry_blk;
298 struct f2fs_dir_entry *de;
299 void *kaddr;
300
301 dentry_page = get_new_data_page(inode, page, 0, true);
302 if (IS_ERR(dentry_page))
303 return PTR_ERR(dentry_page);
304
305 kaddr = kmap_atomic(dentry_page);
306 dentry_blk = (struct f2fs_dentry_block *)kaddr;
307
308 de = &dentry_blk->dentry[0];
309 de->name_len = cpu_to_le16(1);
310 de->hash_code = 0;
311 de->ino = cpu_to_le32(inode->i_ino);
312 memcpy(dentry_blk->filename[0], ".", 1);
313 set_de_type(de, inode);
314
315 de = &dentry_blk->dentry[1];
316 de->hash_code = 0;
317 de->name_len = cpu_to_le16(2);
318 de->ino = cpu_to_le32(parent->i_ino);
319 memcpy(dentry_blk->filename[1], "..", 2);
320 set_de_type(de, inode);
321
322 test_and_set_bit_le(0, &dentry_blk->dentry_bitmap);
323 test_and_set_bit_le(1, &dentry_blk->dentry_bitmap);
324 kunmap_atomic(kaddr);
325
326 set_page_dirty(dentry_page);
327 f2fs_put_page(dentry_page, 1);
328 return 0;
329}
330
331static struct page *init_inode_metadata(struct inode *inode,
332 struct inode *dir, const struct qstr *name)
333{
334 struct page *page;
335 int err;
336
337 if (is_inode_flag_set(F2FS_I(inode), FI_NEW_INODE)) {
338 page = new_inode_page(inode, name);
339 if (IS_ERR(page))
340 return page;
341
342 if (S_ISDIR(inode->i_mode)) {
343 err = make_empty_dir(inode, dir, page);
344 if (err)
345 goto error;
346 }
347
348 err = f2fs_init_acl(inode, dir, page);
349 if (err)
350 goto put_error;
351
352 err = f2fs_init_security(inode, dir, name, page);
353 if (err)
354 goto put_error;
355 } else {
356 page = get_node_page(F2FS_SB(dir->i_sb), inode->i_ino);
357 if (IS_ERR(page))
358 return page;
359
360 set_cold_node(inode, page);
361 }
362
363 init_dent_inode(name, page);
364
365 /*
366 * This file should be checkpointed during fsync.
367 * We lost i_pino from now on.
368 */
369 if (is_inode_flag_set(F2FS_I(inode), FI_INC_LINK)) {
370 file_lost_pino(inode);
371 inc_nlink(inode);
372 }
373 return page;
374
375put_error:
376 f2fs_put_page(page, 1);
377 /* once the failed inode becomes a bad inode, i_mode is S_IFREG */
378 truncate_inode_pages(&inode->i_data, 0);
379 truncate_blocks(inode, 0);
380 remove_dirty_dir_inode(inode);
381error:
382 remove_inode_page(inode);
383 return ERR_PTR(err);
384}
385
386static void update_parent_metadata(struct inode *dir, struct inode *inode,
387 unsigned int current_depth)
388{
389 if (is_inode_flag_set(F2FS_I(inode), FI_NEW_INODE)) {
390 if (S_ISDIR(inode->i_mode)) {
391 inc_nlink(dir);
392 set_inode_flag(F2FS_I(dir), FI_UPDATE_DIR);
393 }
394 clear_inode_flag(F2FS_I(inode), FI_NEW_INODE);
395 }
396 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
397 mark_inode_dirty(dir);
398
399 if (F2FS_I(dir)->i_current_depth != current_depth) {
400 F2FS_I(dir)->i_current_depth = current_depth;
401 set_inode_flag(F2FS_I(dir), FI_UPDATE_DIR);
402 }
403
404 if (is_inode_flag_set(F2FS_I(inode), FI_INC_LINK))
405 clear_inode_flag(F2FS_I(inode), FI_INC_LINK);
406}
407
408static int room_for_filename(struct f2fs_dentry_block *dentry_blk, int slots)
409{
410 int bit_start = 0;
411 int zero_start, zero_end;
412next:
413 zero_start = find_next_zero_bit_le(&dentry_blk->dentry_bitmap,
414 NR_DENTRY_IN_BLOCK,
415 bit_start);
416 if (zero_start >= NR_DENTRY_IN_BLOCK)
417 return NR_DENTRY_IN_BLOCK;
418
419 zero_end = find_next_bit_le(&dentry_blk->dentry_bitmap,
420 NR_DENTRY_IN_BLOCK,
421 zero_start);
422 if (zero_end - zero_start >= slots)
423 return zero_start;
424
425 bit_start = zero_end + 1;
426
427 if (zero_end + 1 >= NR_DENTRY_IN_BLOCK)
428 return NR_DENTRY_IN_BLOCK;
429 goto next;
430}
431
432/*
433 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
434 * f2fs_unlock_op().
435 */
436int __f2fs_add_link(struct inode *dir, const struct qstr *name,
437 struct inode *inode)
438{
439 unsigned int bit_pos;
440 unsigned int level;
441 unsigned int current_depth;
442 unsigned long bidx, block;
443 f2fs_hash_t dentry_hash;
444 struct f2fs_dir_entry *de;
445 unsigned int nbucket, nblock;
446 size_t namelen = name->len;
447 struct page *dentry_page = NULL;
448 struct f2fs_dentry_block *dentry_blk = NULL;
449 int slots = GET_DENTRY_SLOTS(namelen);
450 struct page *page;
451 int err = 0;
452 int i;
453
454 dentry_hash = f2fs_dentry_hash(name->name, name->len);
455 level = 0;
456 current_depth = F2FS_I(dir)->i_current_depth;
457 if (F2FS_I(dir)->chash == dentry_hash) {
458 level = F2FS_I(dir)->clevel;
459 F2FS_I(dir)->chash = 0;
460 }
461
462start:
463 if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
464 return -ENOSPC;
465
466 /* Increase the depth, if required */
467 if (level == current_depth)
468 ++current_depth;
469
470 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
471 nblock = bucket_blocks(level);
472
473 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
474 (le32_to_cpu(dentry_hash) % nbucket));
475
476 for (block = bidx; block <= (bidx + nblock - 1); block++) {
477 dentry_page = get_new_data_page(dir, NULL, block, true);
478 if (IS_ERR(dentry_page))
479 return PTR_ERR(dentry_page);
480
481 dentry_blk = kmap(dentry_page);
482 bit_pos = room_for_filename(dentry_blk, slots);
483 if (bit_pos < NR_DENTRY_IN_BLOCK)
484 goto add_dentry;
485
486 kunmap(dentry_page);
487 f2fs_put_page(dentry_page, 1);
488 }
489
490 /* Move to next level to find the empty slot for new dentry */
491 ++level;
492 goto start;
493add_dentry:
494 f2fs_wait_on_page_writeback(dentry_page, DATA);
495
496 down_write(&F2FS_I(inode)->i_sem);
497 page = init_inode_metadata(inode, dir, name);
498 if (IS_ERR(page)) {
499 err = PTR_ERR(page);
500 goto fail;
501 }
502 de = &dentry_blk->dentry[bit_pos];
503 de->hash_code = dentry_hash;
504 de->name_len = cpu_to_le16(namelen);
505 memcpy(dentry_blk->filename[bit_pos], name->name, name->len);
506 de->ino = cpu_to_le32(inode->i_ino);
507 set_de_type(de, inode);
508 for (i = 0; i < slots; i++)
509 test_and_set_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
510 set_page_dirty(dentry_page);
511
512 /* we don't need to mark_inode_dirty now */
513 F2FS_I(inode)->i_pino = dir->i_ino;
514 update_inode(inode, page);
515 f2fs_put_page(page, 1);
516
517 update_parent_metadata(dir, inode, current_depth);
518fail:
519 up_write(&F2FS_I(inode)->i_sem);
520
521 if (is_inode_flag_set(F2FS_I(dir), FI_UPDATE_DIR)) {
522 update_inode_page(dir);
523 clear_inode_flag(F2FS_I(dir), FI_UPDATE_DIR);
524 }
525 kunmap(dentry_page);
526 f2fs_put_page(dentry_page, 1);
527 return err;
528}
529
530/*
531 * It only removes the dentry from the dentry page,corresponding name
532 * entry in name page does not need to be touched during deletion.
533 */
534void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
535 struct inode *inode)
536{
537 struct f2fs_dentry_block *dentry_blk;
538 unsigned int bit_pos;
539 struct address_space *mapping = page->mapping;
540 struct inode *dir = mapping->host;
541 int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
542 void *kaddr = page_address(page);
543 int i;
544
545 lock_page(page);
546 f2fs_wait_on_page_writeback(page, DATA);
547
548 dentry_blk = (struct f2fs_dentry_block *)kaddr;
549 bit_pos = dentry - (struct f2fs_dir_entry *)dentry_blk->dentry;
550 for (i = 0; i < slots; i++)
551 test_and_clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
552
553 /* Let's check and deallocate this dentry page */
554 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
555 NR_DENTRY_IN_BLOCK,
556 0);
557 kunmap(page); /* kunmap - pair of f2fs_find_entry */
558 set_page_dirty(page);
559
560 dir->i_ctime = dir->i_mtime = CURRENT_TIME;
561
562 if (inode) {
563 struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
564
565 down_write(&F2FS_I(inode)->i_sem);
566
567 if (S_ISDIR(inode->i_mode)) {
568 drop_nlink(dir);
569 update_inode_page(dir);
570 }
571 inode->i_ctime = CURRENT_TIME;
572 drop_nlink(inode);
573 if (S_ISDIR(inode->i_mode)) {
574 drop_nlink(inode);
575 i_size_write(inode, 0);
576 }
577 up_write(&F2FS_I(inode)->i_sem);
578 update_inode_page(inode);
579
580 if (inode->i_nlink == 0)
581 add_orphan_inode(sbi, inode->i_ino);
582 else
583 release_orphan_inode(sbi);
584 }
585
586 if (bit_pos == NR_DENTRY_IN_BLOCK) {
587 truncate_hole(dir, page->index, page->index + 1);
588 clear_page_dirty_for_io(page);
589 ClearPageUptodate(page);
590 inode_dec_dirty_dents(dir);
591 }
592 f2fs_put_page(page, 1);
593}
594
595bool f2fs_empty_dir(struct inode *dir)
596{
597 unsigned long bidx;
598 struct page *dentry_page;
599 unsigned int bit_pos;
600 struct f2fs_dentry_block *dentry_blk;
601 unsigned long nblock = dir_blocks(dir);
602
603 for (bidx = 0; bidx < nblock; bidx++) {
604 void *kaddr;
605 dentry_page = get_lock_data_page(dir, bidx);
606 if (IS_ERR(dentry_page)) {
607 if (PTR_ERR(dentry_page) == -ENOENT)
608 continue;
609 else
610 return false;
611 }
612
613 kaddr = kmap_atomic(dentry_page);
614 dentry_blk = (struct f2fs_dentry_block *)kaddr;
615 if (bidx == 0)
616 bit_pos = 2;
617 else
618 bit_pos = 0;
619 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
620 NR_DENTRY_IN_BLOCK,
621 bit_pos);
622 kunmap_atomic(kaddr);
623
624 f2fs_put_page(dentry_page, 1);
625
626 if (bit_pos < NR_DENTRY_IN_BLOCK)
627 return false;
628 }
629 return true;
630}
631
632static int f2fs_readdir(struct file *file, struct dir_context *ctx)
633{
634 struct inode *inode = file_inode(file);
635 unsigned long npages = dir_blocks(inode);
636 unsigned int bit_pos = 0;
637 struct f2fs_dentry_block *dentry_blk = NULL;
638 struct f2fs_dir_entry *de = NULL;
639 struct page *dentry_page = NULL;
640 unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
641 unsigned char d_type = DT_UNKNOWN;
642
643 bit_pos = ((unsigned long)ctx->pos % NR_DENTRY_IN_BLOCK);
644
645 for (; n < npages; n++) {
646 dentry_page = get_lock_data_page(inode, n);
647 if (IS_ERR(dentry_page))
648 continue;
649
650 dentry_blk = kmap(dentry_page);
651 while (bit_pos < NR_DENTRY_IN_BLOCK) {
652 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
653 NR_DENTRY_IN_BLOCK,
654 bit_pos);
655 if (bit_pos >= NR_DENTRY_IN_BLOCK)
656 break;
657
658 de = &dentry_blk->dentry[bit_pos];
659 if (de->file_type < F2FS_FT_MAX)
660 d_type = f2fs_filetype_table[de->file_type];
661 else
662 d_type = DT_UNKNOWN;
663 if (!dir_emit(ctx,
664 dentry_blk->filename[bit_pos],
665 le16_to_cpu(de->name_len),
666 le32_to_cpu(de->ino), d_type))
667 goto stop;
668
669 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
670 ctx->pos = n * NR_DENTRY_IN_BLOCK + bit_pos;
671 }
672 bit_pos = 0;
673 ctx->pos = (n + 1) * NR_DENTRY_IN_BLOCK;
674 kunmap(dentry_page);
675 f2fs_put_page(dentry_page, 1);
676 dentry_page = NULL;
677 }
678stop:
679 if (dentry_page && !IS_ERR(dentry_page)) {
680 kunmap(dentry_page);
681 f2fs_put_page(dentry_page, 1);
682 }
683
684 return 0;
685}
686
687const struct file_operations f2fs_dir_operations = {
688 .llseek = generic_file_llseek,
689 .read = generic_read_dir,
690 .iterate = f2fs_readdir,
691 .fsync = f2fs_sync_file,
692 .unlocked_ioctl = f2fs_ioctl,
693};
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)
20extern struct kmem_cache *f2fs_cf_name_slab;
21#endif
22
23static 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
29static 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
37static 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 @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
46int f2fs_init_casefolded_name(const struct inode *dir,
47 struct f2fs_filename *fname)
48{
49#if IS_ENABLED(CONFIG_UNICODE)
50 struct super_block *sb = dir->i_sb;
51
52 if (IS_CASEFOLDED(dir) &&
53 !is_dot_dotdot(fname->usr_fname->name, fname->usr_fname->len)) {
54 fname->cf_name.name = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
55 GFP_NOFS, false, F2FS_SB(sb));
56 if (!fname->cf_name.name)
57 return -ENOMEM;
58 fname->cf_name.len = utf8_casefold(sb->s_encoding,
59 fname->usr_fname,
60 fname->cf_name.name,
61 F2FS_NAME_LEN);
62 if ((int)fname->cf_name.len <= 0) {
63 kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
64 fname->cf_name.name = NULL;
65 if (sb_has_strict_encoding(sb))
66 return -EINVAL;
67 /* fall back to treating name as opaque byte sequence */
68 }
69 }
70#endif
71 return 0;
72}
73
74static int __f2fs_setup_filename(const struct inode *dir,
75 const struct fscrypt_name *crypt_name,
76 struct f2fs_filename *fname)
77{
78 int err;
79
80 memset(fname, 0, sizeof(*fname));
81
82 fname->usr_fname = crypt_name->usr_fname;
83 fname->disk_name = crypt_name->disk_name;
84#ifdef CONFIG_FS_ENCRYPTION
85 fname->crypto_buf = crypt_name->crypto_buf;
86#endif
87 if (crypt_name->is_nokey_name) {
88 /* hash was decoded from the no-key name */
89 fname->hash = cpu_to_le32(crypt_name->hash);
90 } else {
91 err = f2fs_init_casefolded_name(dir, fname);
92 if (err) {
93 f2fs_free_filename(fname);
94 return err;
95 }
96 f2fs_hash_filename(dir, fname);
97 }
98 return 0;
99}
100
101/*
102 * Prepare to search for @iname in @dir. This is similar to
103 * fscrypt_setup_filename(), but this also handles computing the casefolded name
104 * and the f2fs dirhash if needed, then packing all the information about this
105 * filename up into a 'struct f2fs_filename'.
106 */
107int f2fs_setup_filename(struct inode *dir, const struct qstr *iname,
108 int lookup, struct f2fs_filename *fname)
109{
110 struct fscrypt_name crypt_name;
111 int err;
112
113 err = fscrypt_setup_filename(dir, iname, lookup, &crypt_name);
114 if (err)
115 return err;
116
117 return __f2fs_setup_filename(dir, &crypt_name, fname);
118}
119
120/*
121 * Prepare to look up @dentry in @dir. This is similar to
122 * fscrypt_prepare_lookup(), but this also handles computing the casefolded name
123 * and the f2fs dirhash if needed, then packing all the information about this
124 * filename up into a 'struct f2fs_filename'.
125 */
126int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
127 struct f2fs_filename *fname)
128{
129 struct fscrypt_name crypt_name;
130 int err;
131
132 err = fscrypt_prepare_lookup(dir, dentry, &crypt_name);
133 if (err)
134 return err;
135
136 return __f2fs_setup_filename(dir, &crypt_name, fname);
137}
138
139void f2fs_free_filename(struct f2fs_filename *fname)
140{
141#ifdef CONFIG_FS_ENCRYPTION
142 kfree(fname->crypto_buf.name);
143 fname->crypto_buf.name = NULL;
144#endif
145#if IS_ENABLED(CONFIG_UNICODE)
146 if (fname->cf_name.name) {
147 kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
148 fname->cf_name.name = NULL;
149 }
150#endif
151}
152
153static unsigned long dir_block_index(unsigned int level,
154 int dir_level, unsigned int idx)
155{
156 unsigned long i;
157 unsigned long bidx = 0;
158
159 for (i = 0; i < level; i++)
160 bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
161 bidx += idx * bucket_blocks(level);
162 return bidx;
163}
164
165static struct f2fs_dir_entry *find_in_block(struct inode *dir,
166 struct page *dentry_page,
167 const struct f2fs_filename *fname,
168 int *max_slots)
169{
170 struct f2fs_dentry_block *dentry_blk;
171 struct f2fs_dentry_ptr d;
172
173 dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
174
175 make_dentry_ptr_block(dir, &d, dentry_blk);
176 return f2fs_find_target_dentry(&d, fname, max_slots);
177}
178
179#if IS_ENABLED(CONFIG_UNICODE)
180/*
181 * Test whether a case-insensitive directory entry matches the filename
182 * being searched for.
183 *
184 * Returns 1 for a match, 0 for no match, and -errno on an error.
185 */
186static int f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
187 const u8 *de_name, u32 de_name_len)
188{
189 const struct super_block *sb = dir->i_sb;
190 const struct unicode_map *um = sb->s_encoding;
191 struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
192 struct qstr entry = QSTR_INIT(de_name, de_name_len);
193 int res;
194
195 if (IS_ENCRYPTED(dir)) {
196 const struct fscrypt_str encrypted_name =
197 FSTR_INIT((u8 *)de_name, de_name_len);
198
199 if (WARN_ON_ONCE(!fscrypt_has_encryption_key(dir)))
200 return -EINVAL;
201
202 decrypted_name.name = kmalloc(de_name_len, GFP_KERNEL);
203 if (!decrypted_name.name)
204 return -ENOMEM;
205 res = fscrypt_fname_disk_to_usr(dir, 0, 0, &encrypted_name,
206 &decrypted_name);
207 if (res < 0)
208 goto out;
209 entry.name = decrypted_name.name;
210 entry.len = decrypted_name.len;
211 }
212
213 res = utf8_strncasecmp_folded(um, name, &entry);
214 /*
215 * In strict mode, ignore invalid names. In non-strict mode,
216 * fall back to treating them as opaque byte sequences.
217 */
218 if (res < 0 && !sb_has_strict_encoding(sb)) {
219 res = name->len == entry.len &&
220 memcmp(name->name, entry.name, name->len) == 0;
221 } else {
222 /* utf8_strncasecmp_folded returns 0 on match */
223 res = (res == 0);
224 }
225out:
226 kfree(decrypted_name.name);
227 return res;
228}
229#endif /* CONFIG_UNICODE */
230
231static inline int f2fs_match_name(const struct inode *dir,
232 const struct f2fs_filename *fname,
233 const u8 *de_name, u32 de_name_len)
234{
235 struct fscrypt_name f;
236
237#if IS_ENABLED(CONFIG_UNICODE)
238 if (fname->cf_name.name) {
239 struct qstr cf = FSTR_TO_QSTR(&fname->cf_name);
240
241 return f2fs_match_ci_name(dir, &cf, de_name, de_name_len);
242 }
243#endif
244 f.usr_fname = fname->usr_fname;
245 f.disk_name = fname->disk_name;
246#ifdef CONFIG_FS_ENCRYPTION
247 f.crypto_buf = fname->crypto_buf;
248#endif
249 return fscrypt_match_name(&f, de_name, de_name_len);
250}
251
252struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
253 const struct f2fs_filename *fname, int *max_slots)
254{
255 struct f2fs_dir_entry *de;
256 unsigned long bit_pos = 0;
257 int max_len = 0;
258 int res = 0;
259
260 if (max_slots)
261 *max_slots = 0;
262 while (bit_pos < d->max) {
263 if (!test_bit_le(bit_pos, d->bitmap)) {
264 bit_pos++;
265 max_len++;
266 continue;
267 }
268
269 de = &d->dentry[bit_pos];
270
271 if (unlikely(!de->name_len)) {
272 bit_pos++;
273 continue;
274 }
275
276 if (de->hash_code == fname->hash) {
277 res = f2fs_match_name(d->inode, fname,
278 d->filename[bit_pos],
279 le16_to_cpu(de->name_len));
280 if (res < 0)
281 return ERR_PTR(res);
282 if (res)
283 goto found;
284 }
285
286 if (max_slots && max_len > *max_slots)
287 *max_slots = max_len;
288 max_len = 0;
289
290 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
291 }
292
293 de = NULL;
294found:
295 if (max_slots && max_len > *max_slots)
296 *max_slots = max_len;
297 return de;
298}
299
300static struct f2fs_dir_entry *find_in_level(struct inode *dir,
301 unsigned int level,
302 const struct f2fs_filename *fname,
303 struct page **res_page)
304{
305 int s = GET_DENTRY_SLOTS(fname->disk_name.len);
306 unsigned int nbucket, nblock;
307 unsigned int bidx, end_block;
308 struct page *dentry_page;
309 struct f2fs_dir_entry *de = NULL;
310 pgoff_t next_pgofs;
311 bool room = false;
312 int max_slots;
313
314 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
315 nblock = bucket_blocks(level);
316
317 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
318 le32_to_cpu(fname->hash) % nbucket);
319 end_block = bidx + nblock;
320
321 while (bidx < end_block) {
322 /* no need to allocate new dentry pages to all the indices */
323 dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
324 if (IS_ERR(dentry_page)) {
325 if (PTR_ERR(dentry_page) == -ENOENT) {
326 room = true;
327 bidx = next_pgofs;
328 continue;
329 } else {
330 *res_page = dentry_page;
331 break;
332 }
333 }
334
335 de = find_in_block(dir, dentry_page, fname, &max_slots);
336 if (IS_ERR(de)) {
337 *res_page = ERR_CAST(de);
338 de = NULL;
339 break;
340 } else if (de) {
341 *res_page = dentry_page;
342 break;
343 }
344
345 if (max_slots >= s)
346 room = true;
347 f2fs_put_page(dentry_page, 0);
348
349 bidx++;
350 }
351
352 if (!de && room && F2FS_I(dir)->chash != fname->hash) {
353 F2FS_I(dir)->chash = fname->hash;
354 F2FS_I(dir)->clevel = level;
355 }
356
357 return de;
358}
359
360struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
361 const struct f2fs_filename *fname,
362 struct page **res_page)
363{
364 unsigned long npages = dir_blocks(dir);
365 struct f2fs_dir_entry *de = NULL;
366 unsigned int max_depth;
367 unsigned int level;
368
369 *res_page = NULL;
370
371 if (f2fs_has_inline_dentry(dir)) {
372 de = f2fs_find_in_inline_dir(dir, fname, res_page);
373 goto out;
374 }
375
376 if (npages == 0)
377 goto out;
378
379 max_depth = F2FS_I(dir)->i_current_depth;
380 if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
381 f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
382 dir->i_ino, max_depth);
383 max_depth = MAX_DIR_HASH_DEPTH;
384 f2fs_i_depth_write(dir, max_depth);
385 }
386
387 for (level = 0; level < max_depth; level++) {
388 de = find_in_level(dir, level, fname, res_page);
389 if (de || IS_ERR(*res_page))
390 break;
391 }
392out:
393 /* This is to increase the speed of f2fs_create */
394 if (!de)
395 F2FS_I(dir)->task = current;
396 return de;
397}
398
399/*
400 * Find an entry in the specified directory with the wanted name.
401 * It returns the page where the entry was found (as a parameter - res_page),
402 * and the entry itself. Page is returned mapped and unlocked.
403 * Entry is guaranteed to be valid.
404 */
405struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
406 const struct qstr *child, struct page **res_page)
407{
408 struct f2fs_dir_entry *de = NULL;
409 struct f2fs_filename fname;
410 int err;
411
412 err = f2fs_setup_filename(dir, child, 1, &fname);
413 if (err) {
414 if (err == -ENOENT)
415 *res_page = NULL;
416 else
417 *res_page = ERR_PTR(err);
418 return NULL;
419 }
420
421 de = __f2fs_find_entry(dir, &fname, res_page);
422
423 f2fs_free_filename(&fname);
424 return de;
425}
426
427struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
428{
429 return f2fs_find_entry(dir, &dotdot_name, p);
430}
431
432ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
433 struct page **page)
434{
435 ino_t res = 0;
436 struct f2fs_dir_entry *de;
437
438 de = f2fs_find_entry(dir, qstr, page);
439 if (de) {
440 res = le32_to_cpu(de->ino);
441 f2fs_put_page(*page, 0);
442 }
443
444 return res;
445}
446
447void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
448 struct page *page, struct inode *inode)
449{
450 enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
451
452 lock_page(page);
453 f2fs_wait_on_page_writeback(page, type, true, true);
454 de->ino = cpu_to_le32(inode->i_ino);
455 de->file_type = fs_umode_to_ftype(inode->i_mode);
456 set_page_dirty(page);
457
458 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
459 f2fs_mark_inode_dirty_sync(dir, false);
460 f2fs_put_page(page, 1);
461}
462
463static void init_dent_inode(struct inode *dir, struct inode *inode,
464 const struct f2fs_filename *fname,
465 struct page *ipage)
466{
467 struct f2fs_inode *ri;
468
469 if (!fname) /* tmpfile case? */
470 return;
471
472 f2fs_wait_on_page_writeback(ipage, NODE, true, true);
473
474 /* copy name info. to this inode page */
475 ri = F2FS_INODE(ipage);
476 ri->i_namelen = cpu_to_le32(fname->disk_name.len);
477 memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
478 if (IS_ENCRYPTED(dir)) {
479 file_set_enc_name(inode);
480 /*
481 * Roll-forward recovery doesn't have encryption keys available,
482 * so it can't compute the dirhash for encrypted+casefolded
483 * filenames. Append it to i_name if possible. Else, disable
484 * roll-forward recovery of the dentry (i.e., make fsync'ing the
485 * file force a checkpoint) by setting LOST_PINO.
486 */
487 if (IS_CASEFOLDED(dir)) {
488 if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
489 F2FS_NAME_LEN)
490 put_unaligned(fname->hash, (f2fs_hash_t *)
491 &ri->i_name[fname->disk_name.len]);
492 else
493 file_lost_pino(inode);
494 }
495 }
496 set_page_dirty(ipage);
497}
498
499void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
500 struct f2fs_dentry_ptr *d)
501{
502 struct fscrypt_str dot = FSTR_INIT(".", 1);
503 struct fscrypt_str dotdot = FSTR_INIT("..", 2);
504
505 /* update dirent of "." */
506 f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
507
508 /* update dirent of ".." */
509 f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
510}
511
512static int make_empty_dir(struct inode *inode,
513 struct inode *parent, struct page *page)
514{
515 struct page *dentry_page;
516 struct f2fs_dentry_block *dentry_blk;
517 struct f2fs_dentry_ptr d;
518
519 if (f2fs_has_inline_dentry(inode))
520 return f2fs_make_empty_inline_dir(inode, parent, page);
521
522 dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
523 if (IS_ERR(dentry_page))
524 return PTR_ERR(dentry_page);
525
526 dentry_blk = page_address(dentry_page);
527
528 make_dentry_ptr_block(NULL, &d, dentry_blk);
529 f2fs_do_make_empty_dir(inode, parent, &d);
530
531 set_page_dirty(dentry_page);
532 f2fs_put_page(dentry_page, 1);
533 return 0;
534}
535
536struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
537 const struct f2fs_filename *fname, struct page *dpage)
538{
539 struct page *page;
540 int err;
541
542 if (is_inode_flag_set(inode, FI_NEW_INODE)) {
543 page = f2fs_new_inode_page(inode);
544 if (IS_ERR(page))
545 return page;
546
547 if (S_ISDIR(inode->i_mode)) {
548 /* in order to handle error case */
549 get_page(page);
550 err = make_empty_dir(inode, dir, page);
551 if (err) {
552 lock_page(page);
553 goto put_error;
554 }
555 put_page(page);
556 }
557
558 err = f2fs_init_acl(inode, dir, page, dpage);
559 if (err)
560 goto put_error;
561
562 err = f2fs_init_security(inode, dir,
563 fname ? fname->usr_fname : NULL, page);
564 if (err)
565 goto put_error;
566
567 if (IS_ENCRYPTED(inode)) {
568 err = fscrypt_set_context(inode, page);
569 if (err)
570 goto put_error;
571 }
572 } else {
573 page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
574 if (IS_ERR(page))
575 return page;
576 }
577
578 init_dent_inode(dir, inode, fname, page);
579
580 /*
581 * This file should be checkpointed during fsync.
582 * We lost i_pino from now on.
583 */
584 if (is_inode_flag_set(inode, FI_INC_LINK)) {
585 if (!S_ISDIR(inode->i_mode))
586 file_lost_pino(inode);
587 /*
588 * If link the tmpfile to alias through linkat path,
589 * we should remove this inode from orphan list.
590 */
591 if (inode->i_nlink == 0)
592 f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
593 f2fs_i_links_write(inode, true);
594 }
595 return page;
596
597put_error:
598 clear_nlink(inode);
599 f2fs_update_inode(inode, page);
600 f2fs_put_page(page, 1);
601 return ERR_PTR(err);
602}
603
604void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
605 unsigned int current_depth)
606{
607 if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
608 if (S_ISDIR(inode->i_mode))
609 f2fs_i_links_write(dir, true);
610 clear_inode_flag(inode, FI_NEW_INODE);
611 }
612 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
613 f2fs_mark_inode_dirty_sync(dir, false);
614
615 if (F2FS_I(dir)->i_current_depth != current_depth)
616 f2fs_i_depth_write(dir, current_depth);
617
618 if (inode && is_inode_flag_set(inode, FI_INC_LINK))
619 clear_inode_flag(inode, FI_INC_LINK);
620}
621
622int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
623{
624 int bit_start = 0;
625 int zero_start, zero_end;
626next:
627 zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
628 if (zero_start >= max_slots)
629 return max_slots;
630
631 zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
632 if (zero_end - zero_start >= slots)
633 return zero_start;
634
635 bit_start = zero_end + 1;
636
637 if (zero_end + 1 >= max_slots)
638 return max_slots;
639 goto next;
640}
641
642bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
643 const struct f2fs_filename *fname)
644{
645 struct f2fs_dentry_ptr d;
646 unsigned int bit_pos;
647 int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
648
649 make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
650
651 bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
652
653 return bit_pos < d.max;
654}
655
656void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
657 const struct fscrypt_str *name, f2fs_hash_t name_hash,
658 unsigned int bit_pos)
659{
660 struct f2fs_dir_entry *de;
661 int slots = GET_DENTRY_SLOTS(name->len);
662 int i;
663
664 de = &d->dentry[bit_pos];
665 de->hash_code = name_hash;
666 de->name_len = cpu_to_le16(name->len);
667 memcpy(d->filename[bit_pos], name->name, name->len);
668 de->ino = cpu_to_le32(ino);
669 de->file_type = fs_umode_to_ftype(mode);
670 for (i = 0; i < slots; i++) {
671 __set_bit_le(bit_pos + i, (void *)d->bitmap);
672 /* avoid wrong garbage data for readdir */
673 if (i)
674 (de + i)->name_len = 0;
675 }
676}
677
678int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
679 struct inode *inode, nid_t ino, umode_t mode)
680{
681 unsigned int bit_pos;
682 unsigned int level;
683 unsigned int current_depth;
684 unsigned long bidx, block;
685 unsigned int nbucket, nblock;
686 struct page *dentry_page = NULL;
687 struct f2fs_dentry_block *dentry_blk = NULL;
688 struct f2fs_dentry_ptr d;
689 struct page *page = NULL;
690 int slots, err = 0;
691
692 level = 0;
693 slots = GET_DENTRY_SLOTS(fname->disk_name.len);
694
695 current_depth = F2FS_I(dir)->i_current_depth;
696 if (F2FS_I(dir)->chash == fname->hash) {
697 level = F2FS_I(dir)->clevel;
698 F2FS_I(dir)->chash = 0;
699 }
700
701start:
702 if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH))
703 return -ENOSPC;
704
705 if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
706 return -ENOSPC;
707
708 /* Increase the depth, if required */
709 if (level == current_depth)
710 ++current_depth;
711
712 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
713 nblock = bucket_blocks(level);
714
715 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
716 (le32_to_cpu(fname->hash) % nbucket));
717
718 for (block = bidx; block <= (bidx + nblock - 1); block++) {
719 dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
720 if (IS_ERR(dentry_page))
721 return PTR_ERR(dentry_page);
722
723 dentry_blk = page_address(dentry_page);
724 bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
725 slots, NR_DENTRY_IN_BLOCK);
726 if (bit_pos < NR_DENTRY_IN_BLOCK)
727 goto add_dentry;
728
729 f2fs_put_page(dentry_page, 1);
730 }
731
732 /* Move to next level to find the empty slot for new dentry */
733 ++level;
734 goto start;
735add_dentry:
736 f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
737
738 if (inode) {
739 f2fs_down_write(&F2FS_I(inode)->i_sem);
740 page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
741 if (IS_ERR(page)) {
742 err = PTR_ERR(page);
743 goto fail;
744 }
745 }
746
747 make_dentry_ptr_block(NULL, &d, dentry_blk);
748 f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
749 bit_pos);
750
751 set_page_dirty(dentry_page);
752
753 if (inode) {
754 f2fs_i_pino_write(inode, dir->i_ino);
755
756 /* synchronize inode page's data from inode cache */
757 if (is_inode_flag_set(inode, FI_NEW_INODE))
758 f2fs_update_inode(inode, page);
759
760 f2fs_put_page(page, 1);
761 }
762
763 f2fs_update_parent_metadata(dir, inode, current_depth);
764fail:
765 if (inode)
766 f2fs_up_write(&F2FS_I(inode)->i_sem);
767
768 f2fs_put_page(dentry_page, 1);
769
770 return err;
771}
772
773int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
774 struct inode *inode, nid_t ino, umode_t mode)
775{
776 int err = -EAGAIN;
777
778 if (f2fs_has_inline_dentry(dir)) {
779 /*
780 * Should get i_xattr_sem to keep the lock order:
781 * i_xattr_sem -> inode_page lock used by f2fs_setxattr.
782 */
783 f2fs_down_read(&F2FS_I(dir)->i_xattr_sem);
784 err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
785 f2fs_up_read(&F2FS_I(dir)->i_xattr_sem);
786 }
787 if (err == -EAGAIN)
788 err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
789
790 f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
791 return err;
792}
793
794/*
795 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
796 * f2fs_unlock_op().
797 */
798int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
799 struct inode *inode, nid_t ino, umode_t mode)
800{
801 struct f2fs_filename fname;
802 struct page *page = NULL;
803 struct f2fs_dir_entry *de = NULL;
804 int err;
805
806 err = f2fs_setup_filename(dir, name, 0, &fname);
807 if (err)
808 return err;
809
810 /*
811 * An immature stackable filesystem shows a race condition between lookup
812 * and create. If we have same task when doing lookup and create, it's
813 * definitely fine as expected by VFS normally. Otherwise, let's just
814 * verify on-disk dentry one more time, which guarantees filesystem
815 * consistency more.
816 */
817 if (current != F2FS_I(dir)->task) {
818 de = __f2fs_find_entry(dir, &fname, &page);
819 F2FS_I(dir)->task = NULL;
820 }
821 if (de) {
822 f2fs_put_page(page, 0);
823 err = -EEXIST;
824 } else if (IS_ERR(page)) {
825 err = PTR_ERR(page);
826 } else {
827 err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
828 }
829 f2fs_free_filename(&fname);
830 return err;
831}
832
833int f2fs_do_tmpfile(struct inode *inode, struct inode *dir,
834 struct f2fs_filename *fname)
835{
836 struct page *page;
837 int err = 0;
838
839 f2fs_down_write(&F2FS_I(inode)->i_sem);
840 page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
841 if (IS_ERR(page)) {
842 err = PTR_ERR(page);
843 goto fail;
844 }
845 f2fs_put_page(page, 1);
846
847 clear_inode_flag(inode, FI_NEW_INODE);
848 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
849fail:
850 f2fs_up_write(&F2FS_I(inode)->i_sem);
851 return err;
852}
853
854void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
855{
856 struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
857
858 f2fs_down_write(&F2FS_I(inode)->i_sem);
859
860 if (S_ISDIR(inode->i_mode))
861 f2fs_i_links_write(dir, false);
862 inode_set_ctime_current(inode);
863
864 f2fs_i_links_write(inode, false);
865 if (S_ISDIR(inode->i_mode)) {
866 f2fs_i_links_write(inode, false);
867 f2fs_i_size_write(inode, 0);
868 }
869 f2fs_up_write(&F2FS_I(inode)->i_sem);
870
871 if (inode->i_nlink == 0)
872 f2fs_add_orphan_inode(inode);
873 else
874 f2fs_release_orphan_inode(sbi);
875}
876
877/*
878 * It only removes the dentry from the dentry page, corresponding name
879 * entry in name page does not need to be touched during deletion.
880 */
881void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
882 struct inode *dir, struct inode *inode)
883{
884 struct f2fs_dentry_block *dentry_blk;
885 unsigned int bit_pos;
886 int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
887 int i;
888
889 f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
890
891 if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
892 f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
893
894 if (f2fs_has_inline_dentry(dir))
895 return f2fs_delete_inline_entry(dentry, page, dir, inode);
896
897 lock_page(page);
898 f2fs_wait_on_page_writeback(page, DATA, true, true);
899
900 dentry_blk = page_address(page);
901 bit_pos = dentry - dentry_blk->dentry;
902 for (i = 0; i < slots; i++)
903 __clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
904
905 /* Let's check and deallocate this dentry page */
906 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
907 NR_DENTRY_IN_BLOCK,
908 0);
909 set_page_dirty(page);
910
911 if (bit_pos == NR_DENTRY_IN_BLOCK &&
912 !f2fs_truncate_hole(dir, page->index, page->index + 1)) {
913 f2fs_clear_page_cache_dirty_tag(page);
914 clear_page_dirty_for_io(page);
915 ClearPageUptodate(page);
916 clear_page_private_all(page);
917
918 inode_dec_dirty_pages(dir);
919 f2fs_remove_dirty_inode(dir);
920 }
921 f2fs_put_page(page, 1);
922
923 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
924 f2fs_mark_inode_dirty_sync(dir, false);
925
926 if (inode)
927 f2fs_drop_nlink(dir, inode);
928}
929
930bool f2fs_empty_dir(struct inode *dir)
931{
932 unsigned long bidx = 0;
933 struct page *dentry_page;
934 unsigned int bit_pos;
935 struct f2fs_dentry_block *dentry_blk;
936 unsigned long nblock = dir_blocks(dir);
937
938 if (f2fs_has_inline_dentry(dir))
939 return f2fs_empty_inline_dir(dir);
940
941 while (bidx < nblock) {
942 pgoff_t next_pgofs;
943
944 dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
945 if (IS_ERR(dentry_page)) {
946 if (PTR_ERR(dentry_page) == -ENOENT) {
947 bidx = next_pgofs;
948 continue;
949 } else {
950 return false;
951 }
952 }
953
954 dentry_blk = page_address(dentry_page);
955 if (bidx == 0)
956 bit_pos = 2;
957 else
958 bit_pos = 0;
959 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
960 NR_DENTRY_IN_BLOCK,
961 bit_pos);
962
963 f2fs_put_page(dentry_page, 0);
964
965 if (bit_pos < NR_DENTRY_IN_BLOCK)
966 return false;
967
968 bidx++;
969 }
970 return true;
971}
972
973int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
974 unsigned int start_pos, struct fscrypt_str *fstr)
975{
976 unsigned char d_type = DT_UNKNOWN;
977 unsigned int bit_pos;
978 struct f2fs_dir_entry *de = NULL;
979 struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
980 struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
981 struct blk_plug plug;
982 bool readdir_ra = sbi->readdir_ra;
983 bool found_valid_dirent = false;
984 int err = 0;
985
986 bit_pos = ((unsigned long)ctx->pos % d->max);
987
988 if (readdir_ra)
989 blk_start_plug(&plug);
990
991 while (bit_pos < d->max) {
992 bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
993 if (bit_pos >= d->max)
994 break;
995
996 de = &d->dentry[bit_pos];
997 if (de->name_len == 0) {
998 if (found_valid_dirent || !bit_pos) {
999 f2fs_warn_ratelimited(sbi,
1000 "invalid namelen(0), ino:%u, run fsck to fix.",
1001 le32_to_cpu(de->ino));
1002 set_sbi_flag(sbi, SBI_NEED_FSCK);
1003 }
1004 bit_pos++;
1005 ctx->pos = start_pos + bit_pos;
1006 continue;
1007 }
1008
1009 d_type = fs_ftype_to_dtype(de->file_type);
1010
1011 de_name.name = d->filename[bit_pos];
1012 de_name.len = le16_to_cpu(de->name_len);
1013
1014 /* check memory boundary before moving forward */
1015 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
1016 if (unlikely(bit_pos > d->max ||
1017 le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
1018 f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
1019 __func__, le16_to_cpu(de->name_len));
1020 set_sbi_flag(sbi, SBI_NEED_FSCK);
1021 err = -EFSCORRUPTED;
1022 f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
1023 goto out;
1024 }
1025
1026 if (IS_ENCRYPTED(d->inode)) {
1027 int save_len = fstr->len;
1028
1029 err = fscrypt_fname_disk_to_usr(d->inode,
1030 (u32)le32_to_cpu(de->hash_code),
1031 0, &de_name, fstr);
1032 if (err)
1033 goto out;
1034
1035 de_name = *fstr;
1036 fstr->len = save_len;
1037 }
1038
1039 if (!dir_emit(ctx, de_name.name, de_name.len,
1040 le32_to_cpu(de->ino), d_type)) {
1041 err = 1;
1042 goto out;
1043 }
1044
1045 if (readdir_ra)
1046 f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1047
1048 ctx->pos = start_pos + bit_pos;
1049 found_valid_dirent = true;
1050 }
1051out:
1052 if (readdir_ra)
1053 blk_finish_plug(&plug);
1054 return err;
1055}
1056
1057static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1058{
1059 struct inode *inode = file_inode(file);
1060 unsigned long npages = dir_blocks(inode);
1061 struct f2fs_dentry_block *dentry_blk = NULL;
1062 struct page *dentry_page = NULL;
1063 struct file_ra_state *ra = &file->f_ra;
1064 loff_t start_pos = ctx->pos;
1065 unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1066 struct f2fs_dentry_ptr d;
1067 struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1068 int err = 0;
1069
1070 if (IS_ENCRYPTED(inode)) {
1071 err = fscrypt_prepare_readdir(inode);
1072 if (err)
1073 goto out;
1074
1075 err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1076 if (err < 0)
1077 goto out;
1078 }
1079
1080 if (f2fs_has_inline_dentry(inode)) {
1081 err = f2fs_read_inline_dir(file, ctx, &fstr);
1082 goto out_free;
1083 }
1084
1085 for (; n < npages; ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1086 pgoff_t next_pgofs;
1087
1088 /* allow readdir() to be interrupted */
1089 if (fatal_signal_pending(current)) {
1090 err = -ERESTARTSYS;
1091 goto out_free;
1092 }
1093 cond_resched();
1094
1095 /* readahead for multi pages of dir */
1096 if (npages - n > 1 && !ra_has_index(ra, n))
1097 page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1098 min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1099
1100 dentry_page = f2fs_find_data_page(inode, n, &next_pgofs);
1101 if (IS_ERR(dentry_page)) {
1102 err = PTR_ERR(dentry_page);
1103 if (err == -ENOENT) {
1104 err = 0;
1105 n = next_pgofs;
1106 continue;
1107 } else {
1108 goto out_free;
1109 }
1110 }
1111
1112 dentry_blk = page_address(dentry_page);
1113
1114 make_dentry_ptr_block(inode, &d, dentry_blk);
1115
1116 err = f2fs_fill_dentries(ctx, &d,
1117 n * NR_DENTRY_IN_BLOCK, &fstr);
1118 if (err) {
1119 f2fs_put_page(dentry_page, 0);
1120 break;
1121 }
1122
1123 f2fs_put_page(dentry_page, 0);
1124
1125 n++;
1126 }
1127out_free:
1128 fscrypt_fname_free_buffer(&fstr);
1129out:
1130 trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1131 return err < 0 ? err : 0;
1132}
1133
1134const struct file_operations f2fs_dir_operations = {
1135 .llseek = generic_file_llseek,
1136 .read = generic_read_dir,
1137 .iterate_shared = f2fs_readdir,
1138 .fsync = f2fs_sync_file,
1139 .unlocked_ioctl = f2fs_ioctl,
1140#ifdef CONFIG_COMPAT
1141 .compat_ioctl = f2fs_compat_ioctl,
1142#endif
1143};