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
dir_blocks(struct inode * inode)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
dir_buckets(unsigned int level,int dir_level)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
bucket_blocks(unsigned int level)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 @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
f2fs_init_casefolded_name(const struct inode * dir,struct f2fs_filename * fname)46 int 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
__f2fs_setup_filename(const struct inode * dir,const struct fscrypt_name * crypt_name,struct f2fs_filename * fname)74 static 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 */
f2fs_setup_filename(struct inode * dir,const struct qstr * iname,int lookup,struct f2fs_filename * fname)107 int 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 */
f2fs_prepare_lookup(struct inode * dir,struct dentry * dentry,struct f2fs_filename * fname)126 int 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
f2fs_free_filename(struct f2fs_filename * fname)139 void 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
dir_block_index(unsigned int level,int dir_level,unsigned int idx)153 static 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 += mul_u32_u32(dir_buckets(i, dir_level),
161 bucket_blocks(i));
162 bidx += idx * bucket_blocks(level);
163 return bidx;
164 }
165
find_in_block(struct inode * dir,struct page * dentry_page,const struct f2fs_filename * fname,int * max_slots,bool use_hash)166 static struct f2fs_dir_entry *find_in_block(struct inode *dir,
167 struct page *dentry_page,
168 const struct f2fs_filename *fname,
169 int *max_slots,
170 bool use_hash)
171 {
172 struct f2fs_dentry_block *dentry_blk;
173 struct f2fs_dentry_ptr d;
174
175 dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
176
177 make_dentry_ptr_block(dir, &d, dentry_blk);
178 return f2fs_find_target_dentry(&d, fname, max_slots, use_hash);
179 }
180
181 #if IS_ENABLED(CONFIG_UNICODE)
182 /*
183 * Test whether a case-insensitive directory entry matches the filename
184 * being searched for.
185 *
186 * Returns 1 for a match, 0 for no match, and -errno on an error.
187 */
f2fs_match_ci_name(const struct inode * dir,const struct qstr * name,const u8 * de_name,u32 de_name_len)188 static int f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
189 const u8 *de_name, u32 de_name_len)
190 {
191 const struct super_block *sb = dir->i_sb;
192 const struct unicode_map *um = sb->s_encoding;
193 struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
194 struct qstr entry = QSTR_INIT(de_name, de_name_len);
195 int res;
196
197 if (IS_ENCRYPTED(dir)) {
198 const struct fscrypt_str encrypted_name =
199 FSTR_INIT((u8 *)de_name, de_name_len);
200
201 if (WARN_ON_ONCE(!fscrypt_has_encryption_key(dir)))
202 return -EINVAL;
203
204 decrypted_name.name = kmalloc(de_name_len, GFP_KERNEL);
205 if (!decrypted_name.name)
206 return -ENOMEM;
207 res = fscrypt_fname_disk_to_usr(dir, 0, 0, &encrypted_name,
208 &decrypted_name);
209 if (res < 0)
210 goto out;
211 entry.name = decrypted_name.name;
212 entry.len = decrypted_name.len;
213 }
214
215 res = utf8_strncasecmp_folded(um, name, &entry);
216 /*
217 * In strict mode, ignore invalid names. In non-strict mode,
218 * fall back to treating them as opaque byte sequences.
219 */
220 if (res < 0 && !sb_has_strict_encoding(sb)) {
221 res = name->len == entry.len &&
222 memcmp(name->name, entry.name, name->len) == 0;
223 } else {
224 /* utf8_strncasecmp_folded returns 0 on match */
225 res = (res == 0);
226 }
227 out:
228 kfree(decrypted_name.name);
229 return res;
230 }
231 #endif /* CONFIG_UNICODE */
232
f2fs_match_name(const struct inode * dir,const struct f2fs_filename * fname,const u8 * de_name,u32 de_name_len)233 static inline int f2fs_match_name(const struct inode *dir,
234 const struct f2fs_filename *fname,
235 const u8 *de_name, u32 de_name_len)
236 {
237 struct fscrypt_name f;
238
239 #if IS_ENABLED(CONFIG_UNICODE)
240 if (fname->cf_name.name) {
241 struct qstr cf = FSTR_TO_QSTR(&fname->cf_name);
242
243 return f2fs_match_ci_name(dir, &cf, de_name, de_name_len);
244 }
245 #endif
246 f.usr_fname = fname->usr_fname;
247 f.disk_name = fname->disk_name;
248 #ifdef CONFIG_FS_ENCRYPTION
249 f.crypto_buf = fname->crypto_buf;
250 #endif
251 return fscrypt_match_name(&f, de_name, de_name_len);
252 }
253
f2fs_find_target_dentry(const struct f2fs_dentry_ptr * d,const struct f2fs_filename * fname,int * max_slots,bool use_hash)254 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
255 const struct f2fs_filename *fname, int *max_slots,
256 bool use_hash)
257 {
258 struct f2fs_dir_entry *de;
259 unsigned long bit_pos = 0;
260 int max_len = 0;
261 int res = 0;
262
263 if (max_slots)
264 *max_slots = 0;
265 while (bit_pos < d->max) {
266 if (!test_bit_le(bit_pos, d->bitmap)) {
267 bit_pos++;
268 max_len++;
269 continue;
270 }
271
272 de = &d->dentry[bit_pos];
273
274 if (unlikely(!de->name_len)) {
275 bit_pos++;
276 continue;
277 }
278
279 if (!use_hash || de->hash_code == fname->hash) {
280 res = f2fs_match_name(d->inode, fname,
281 d->filename[bit_pos],
282 le16_to_cpu(de->name_len));
283 if (res < 0)
284 return ERR_PTR(res);
285 if (res)
286 goto found;
287 }
288
289 if (max_slots && max_len > *max_slots)
290 *max_slots = max_len;
291 max_len = 0;
292
293 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
294 }
295
296 de = NULL;
297 found:
298 if (max_slots && max_len > *max_slots)
299 *max_slots = max_len;
300 return de;
301 }
302
find_in_level(struct inode * dir,unsigned int level,const struct f2fs_filename * fname,struct page ** res_page,bool use_hash)303 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
304 unsigned int level,
305 const struct f2fs_filename *fname,
306 struct page **res_page,
307 bool use_hash)
308 {
309 int s = GET_DENTRY_SLOTS(fname->disk_name.len);
310 unsigned int nbucket, nblock;
311 unsigned int bidx, end_block, bucket_no;
312 struct page *dentry_page;
313 struct f2fs_dir_entry *de = NULL;
314 pgoff_t next_pgofs;
315 bool room = false;
316 int max_slots;
317
318 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
319 nblock = bucket_blocks(level);
320
321 bucket_no = use_hash ? le32_to_cpu(fname->hash) % nbucket : 0;
322
323 start_find_bucket:
324 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
325 bucket_no);
326 end_block = bidx + nblock;
327
328 while (bidx < end_block) {
329 /* no need to allocate new dentry pages to all the indices */
330 dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
331 if (IS_ERR(dentry_page)) {
332 if (PTR_ERR(dentry_page) == -ENOENT) {
333 room = true;
334 bidx = next_pgofs;
335 continue;
336 } else {
337 *res_page = dentry_page;
338 break;
339 }
340 }
341
342 de = find_in_block(dir, dentry_page, fname, &max_slots, use_hash);
343 if (IS_ERR(de)) {
344 *res_page = ERR_CAST(de);
345 de = NULL;
346 break;
347 } else if (de) {
348 *res_page = dentry_page;
349 break;
350 }
351
352 if (max_slots >= s)
353 room = true;
354 f2fs_put_page(dentry_page, 0);
355
356 bidx++;
357 }
358
359 if (de)
360 return de;
361
362 if (likely(use_hash)) {
363 if (room && F2FS_I(dir)->chash != fname->hash) {
364 F2FS_I(dir)->chash = fname->hash;
365 F2FS_I(dir)->clevel = level;
366 }
367 } else if (++bucket_no < nbucket) {
368 goto start_find_bucket;
369 }
370 return NULL;
371 }
372
__f2fs_find_entry(struct inode * dir,const struct f2fs_filename * fname,struct page ** res_page)373 struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
374 const struct f2fs_filename *fname,
375 struct page **res_page)
376 {
377 unsigned long npages = dir_blocks(dir);
378 struct f2fs_dir_entry *de = NULL;
379 unsigned int max_depth;
380 unsigned int level;
381 bool use_hash = true;
382
383 *res_page = NULL;
384
385 #if IS_ENABLED(CONFIG_UNICODE)
386 start_find_entry:
387 #endif
388 if (f2fs_has_inline_dentry(dir)) {
389 de = f2fs_find_in_inline_dir(dir, fname, res_page, use_hash);
390 goto out;
391 }
392
393 if (npages == 0)
394 goto out;
395
396 max_depth = F2FS_I(dir)->i_current_depth;
397 if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
398 f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
399 dir->i_ino, max_depth);
400 max_depth = MAX_DIR_HASH_DEPTH;
401 f2fs_i_depth_write(dir, max_depth);
402 }
403
404 for (level = 0; level < max_depth; level++) {
405 de = find_in_level(dir, level, fname, res_page, use_hash);
406 if (de || IS_ERR(*res_page))
407 break;
408 }
409
410 out:
411 #if IS_ENABLED(CONFIG_UNICODE)
412 if (IS_CASEFOLDED(dir) && !de && use_hash) {
413 use_hash = false;
414 goto start_find_entry;
415 }
416 #endif
417 /* This is to increase the speed of f2fs_create */
418 if (!de)
419 F2FS_I(dir)->task = current;
420 return de;
421 }
422
423 /*
424 * Find an entry in the specified directory with the wanted name.
425 * It returns the page where the entry was found (as a parameter - res_page),
426 * and the entry itself. Page is returned mapped and unlocked.
427 * Entry is guaranteed to be valid.
428 */
f2fs_find_entry(struct inode * dir,const struct qstr * child,struct page ** res_page)429 struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
430 const struct qstr *child, struct page **res_page)
431 {
432 struct f2fs_dir_entry *de = NULL;
433 struct f2fs_filename fname;
434 int err;
435
436 err = f2fs_setup_filename(dir, child, 1, &fname);
437 if (err) {
438 if (err == -ENOENT)
439 *res_page = NULL;
440 else
441 *res_page = ERR_PTR(err);
442 return NULL;
443 }
444
445 de = __f2fs_find_entry(dir, &fname, res_page);
446
447 f2fs_free_filename(&fname);
448 return de;
449 }
450
f2fs_parent_dir(struct inode * dir,struct page ** p)451 struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
452 {
453 return f2fs_find_entry(dir, &dotdot_name, p);
454 }
455
f2fs_inode_by_name(struct inode * dir,const struct qstr * qstr,struct page ** page)456 ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
457 struct page **page)
458 {
459 ino_t res = 0;
460 struct f2fs_dir_entry *de;
461
462 de = f2fs_find_entry(dir, qstr, page);
463 if (de) {
464 res = le32_to_cpu(de->ino);
465 f2fs_put_page(*page, 0);
466 }
467
468 return res;
469 }
470
f2fs_set_link(struct inode * dir,struct f2fs_dir_entry * de,struct page * page,struct inode * inode)471 void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
472 struct page *page, struct inode *inode)
473 {
474 enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
475
476 lock_page(page);
477 f2fs_wait_on_page_writeback(page, type, true, true);
478 de->ino = cpu_to_le32(inode->i_ino);
479 de->file_type = fs_umode_to_ftype(inode->i_mode);
480 set_page_dirty(page);
481
482 dir->i_mtime = inode_set_ctime_current(dir);
483 f2fs_mark_inode_dirty_sync(dir, false);
484 f2fs_put_page(page, 1);
485 }
486
init_dent_inode(struct inode * dir,struct inode * inode,const struct f2fs_filename * fname,struct page * ipage)487 static void init_dent_inode(struct inode *dir, struct inode *inode,
488 const struct f2fs_filename *fname,
489 struct page *ipage)
490 {
491 struct f2fs_inode *ri;
492
493 if (!fname) /* tmpfile case? */
494 return;
495
496 f2fs_wait_on_page_writeback(ipage, NODE, true, true);
497
498 /* copy name info. to this inode page */
499 ri = F2FS_INODE(ipage);
500 ri->i_namelen = cpu_to_le32(fname->disk_name.len);
501 memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
502 if (IS_ENCRYPTED(dir)) {
503 file_set_enc_name(inode);
504 /*
505 * Roll-forward recovery doesn't have encryption keys available,
506 * so it can't compute the dirhash for encrypted+casefolded
507 * filenames. Append it to i_name if possible. Else, disable
508 * roll-forward recovery of the dentry (i.e., make fsync'ing the
509 * file force a checkpoint) by setting LOST_PINO.
510 */
511 if (IS_CASEFOLDED(dir)) {
512 if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
513 F2FS_NAME_LEN)
514 put_unaligned(fname->hash, (f2fs_hash_t *)
515 &ri->i_name[fname->disk_name.len]);
516 else
517 file_lost_pino(inode);
518 }
519 }
520 set_page_dirty(ipage);
521 }
522
f2fs_do_make_empty_dir(struct inode * inode,struct inode * parent,struct f2fs_dentry_ptr * d)523 void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
524 struct f2fs_dentry_ptr *d)
525 {
526 struct fscrypt_str dot = FSTR_INIT(".", 1);
527 struct fscrypt_str dotdot = FSTR_INIT("..", 2);
528
529 /* update dirent of "." */
530 f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
531
532 /* update dirent of ".." */
533 f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
534 }
535
make_empty_dir(struct inode * inode,struct inode * parent,struct page * page)536 static int make_empty_dir(struct inode *inode,
537 struct inode *parent, struct page *page)
538 {
539 struct page *dentry_page;
540 struct f2fs_dentry_block *dentry_blk;
541 struct f2fs_dentry_ptr d;
542
543 if (f2fs_has_inline_dentry(inode))
544 return f2fs_make_empty_inline_dir(inode, parent, page);
545
546 dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
547 if (IS_ERR(dentry_page))
548 return PTR_ERR(dentry_page);
549
550 dentry_blk = page_address(dentry_page);
551
552 make_dentry_ptr_block(NULL, &d, dentry_blk);
553 f2fs_do_make_empty_dir(inode, parent, &d);
554
555 set_page_dirty(dentry_page);
556 f2fs_put_page(dentry_page, 1);
557 return 0;
558 }
559
f2fs_init_inode_metadata(struct inode * inode,struct inode * dir,const struct f2fs_filename * fname,struct page * dpage)560 struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
561 const struct f2fs_filename *fname, struct page *dpage)
562 {
563 struct page *page;
564 int err;
565
566 if (is_inode_flag_set(inode, FI_NEW_INODE)) {
567 page = f2fs_new_inode_page(inode);
568 if (IS_ERR(page))
569 return page;
570
571 if (S_ISDIR(inode->i_mode)) {
572 /* in order to handle error case */
573 get_page(page);
574 err = make_empty_dir(inode, dir, page);
575 if (err) {
576 lock_page(page);
577 goto put_error;
578 }
579 put_page(page);
580 }
581
582 err = f2fs_init_acl(inode, dir, page, dpage);
583 if (err)
584 goto put_error;
585
586 err = f2fs_init_security(inode, dir,
587 fname ? fname->usr_fname : NULL, page);
588 if (err)
589 goto put_error;
590
591 if (IS_ENCRYPTED(inode)) {
592 err = fscrypt_set_context(inode, page);
593 if (err)
594 goto put_error;
595 }
596 } else {
597 page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
598 if (IS_ERR(page))
599 return page;
600 }
601
602 init_dent_inode(dir, inode, fname, page);
603
604 /*
605 * This file should be checkpointed during fsync.
606 * We lost i_pino from now on.
607 */
608 if (is_inode_flag_set(inode, FI_INC_LINK)) {
609 if (!S_ISDIR(inode->i_mode))
610 file_lost_pino(inode);
611 /*
612 * If link the tmpfile to alias through linkat path,
613 * we should remove this inode from orphan list.
614 */
615 if (inode->i_nlink == 0)
616 f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
617 f2fs_i_links_write(inode, true);
618 }
619 return page;
620
621 put_error:
622 clear_nlink(inode);
623 f2fs_update_inode(inode, page);
624 f2fs_put_page(page, 1);
625 return ERR_PTR(err);
626 }
627
f2fs_update_parent_metadata(struct inode * dir,struct inode * inode,unsigned int current_depth)628 void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
629 unsigned int current_depth)
630 {
631 if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
632 if (S_ISDIR(inode->i_mode))
633 f2fs_i_links_write(dir, true);
634 clear_inode_flag(inode, FI_NEW_INODE);
635 }
636 dir->i_mtime = inode_set_ctime_current(dir);
637 f2fs_mark_inode_dirty_sync(dir, false);
638
639 if (F2FS_I(dir)->i_current_depth != current_depth)
640 f2fs_i_depth_write(dir, current_depth);
641
642 if (inode && is_inode_flag_set(inode, FI_INC_LINK))
643 clear_inode_flag(inode, FI_INC_LINK);
644 }
645
f2fs_room_for_filename(const void * bitmap,int slots,int max_slots)646 int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
647 {
648 int bit_start = 0;
649 int zero_start, zero_end;
650 next:
651 zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
652 if (zero_start >= max_slots)
653 return max_slots;
654
655 zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
656 if (zero_end - zero_start >= slots)
657 return zero_start;
658
659 bit_start = zero_end + 1;
660
661 if (zero_end + 1 >= max_slots)
662 return max_slots;
663 goto next;
664 }
665
f2fs_has_enough_room(struct inode * dir,struct page * ipage,const struct f2fs_filename * fname)666 bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
667 const struct f2fs_filename *fname)
668 {
669 struct f2fs_dentry_ptr d;
670 unsigned int bit_pos;
671 int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
672
673 make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
674
675 bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
676
677 return bit_pos < d.max;
678 }
679
f2fs_update_dentry(nid_t ino,umode_t mode,struct f2fs_dentry_ptr * d,const struct fscrypt_str * name,f2fs_hash_t name_hash,unsigned int bit_pos)680 void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
681 const struct fscrypt_str *name, f2fs_hash_t name_hash,
682 unsigned int bit_pos)
683 {
684 struct f2fs_dir_entry *de;
685 int slots = GET_DENTRY_SLOTS(name->len);
686 int i;
687
688 de = &d->dentry[bit_pos];
689 de->hash_code = name_hash;
690 de->name_len = cpu_to_le16(name->len);
691 memcpy(d->filename[bit_pos], name->name, name->len);
692 de->ino = cpu_to_le32(ino);
693 de->file_type = fs_umode_to_ftype(mode);
694 for (i = 0; i < slots; i++) {
695 __set_bit_le(bit_pos + i, (void *)d->bitmap);
696 /* avoid wrong garbage data for readdir */
697 if (i)
698 (de + i)->name_len = 0;
699 }
700 }
701
f2fs_add_regular_entry(struct inode * dir,const struct f2fs_filename * fname,struct inode * inode,nid_t ino,umode_t mode)702 int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
703 struct inode *inode, nid_t ino, umode_t mode)
704 {
705 unsigned int bit_pos;
706 unsigned int level;
707 unsigned int current_depth;
708 unsigned long bidx, block;
709 unsigned int nbucket, nblock;
710 struct page *dentry_page = NULL;
711 struct f2fs_dentry_block *dentry_blk = NULL;
712 struct f2fs_dentry_ptr d;
713 struct page *page = NULL;
714 int slots, err = 0;
715
716 level = 0;
717 slots = GET_DENTRY_SLOTS(fname->disk_name.len);
718
719 current_depth = F2FS_I(dir)->i_current_depth;
720 if (F2FS_I(dir)->chash == fname->hash) {
721 level = F2FS_I(dir)->clevel;
722 F2FS_I(dir)->chash = 0;
723 }
724
725 start:
726 if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH))
727 return -ENOSPC;
728
729 if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
730 return -ENOSPC;
731
732 /* Increase the depth, if required */
733 if (level == current_depth)
734 ++current_depth;
735
736 nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
737 nblock = bucket_blocks(level);
738
739 bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
740 (le32_to_cpu(fname->hash) % nbucket));
741
742 for (block = bidx; block <= (bidx + nblock - 1); block++) {
743 dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
744 if (IS_ERR(dentry_page))
745 return PTR_ERR(dentry_page);
746
747 dentry_blk = page_address(dentry_page);
748 bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
749 slots, NR_DENTRY_IN_BLOCK);
750 if (bit_pos < NR_DENTRY_IN_BLOCK)
751 goto add_dentry;
752
753 f2fs_put_page(dentry_page, 1);
754 }
755
756 /* Move to next level to find the empty slot for new dentry */
757 ++level;
758 goto start;
759 add_dentry:
760 f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
761
762 if (inode) {
763 f2fs_down_write(&F2FS_I(inode)->i_sem);
764 page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
765 if (IS_ERR(page)) {
766 err = PTR_ERR(page);
767 goto fail;
768 }
769 }
770
771 make_dentry_ptr_block(NULL, &d, dentry_blk);
772 f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
773 bit_pos);
774
775 set_page_dirty(dentry_page);
776
777 if (inode) {
778 f2fs_i_pino_write(inode, dir->i_ino);
779
780 /* synchronize inode page's data from inode cache */
781 if (is_inode_flag_set(inode, FI_NEW_INODE))
782 f2fs_update_inode(inode, page);
783
784 f2fs_put_page(page, 1);
785 }
786
787 f2fs_update_parent_metadata(dir, inode, current_depth);
788 fail:
789 if (inode)
790 f2fs_up_write(&F2FS_I(inode)->i_sem);
791
792 f2fs_put_page(dentry_page, 1);
793
794 return err;
795 }
796
f2fs_add_dentry(struct inode * dir,const struct f2fs_filename * fname,struct inode * inode,nid_t ino,umode_t mode)797 int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
798 struct inode *inode, nid_t ino, umode_t mode)
799 {
800 int err = -EAGAIN;
801
802 if (f2fs_has_inline_dentry(dir)) {
803 /*
804 * Should get i_xattr_sem to keep the lock order:
805 * i_xattr_sem -> inode_page lock used by f2fs_setxattr.
806 */
807 f2fs_down_read(&F2FS_I(dir)->i_xattr_sem);
808 err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
809 f2fs_up_read(&F2FS_I(dir)->i_xattr_sem);
810 }
811 if (err == -EAGAIN)
812 err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
813
814 f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
815 return err;
816 }
817
818 /*
819 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
820 * f2fs_unlock_op().
821 */
f2fs_do_add_link(struct inode * dir,const struct qstr * name,struct inode * inode,nid_t ino,umode_t mode)822 int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
823 struct inode *inode, nid_t ino, umode_t mode)
824 {
825 struct f2fs_filename fname;
826 struct page *page = NULL;
827 struct f2fs_dir_entry *de = NULL;
828 int err;
829
830 err = f2fs_setup_filename(dir, name, 0, &fname);
831 if (err)
832 return err;
833
834 /*
835 * An immature stackable filesystem shows a race condition between lookup
836 * and create. If we have same task when doing lookup and create, it's
837 * definitely fine as expected by VFS normally. Otherwise, let's just
838 * verify on-disk dentry one more time, which guarantees filesystem
839 * consistency more.
840 */
841 if (current != F2FS_I(dir)->task) {
842 de = __f2fs_find_entry(dir, &fname, &page);
843 F2FS_I(dir)->task = NULL;
844 }
845 if (de) {
846 f2fs_put_page(page, 0);
847 err = -EEXIST;
848 } else if (IS_ERR(page)) {
849 err = PTR_ERR(page);
850 } else {
851 err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
852 }
853 f2fs_free_filename(&fname);
854 return err;
855 }
856
f2fs_do_tmpfile(struct inode * inode,struct inode * dir,struct f2fs_filename * fname)857 int f2fs_do_tmpfile(struct inode *inode, struct inode *dir,
858 struct f2fs_filename *fname)
859 {
860 struct page *page;
861 int err = 0;
862
863 f2fs_down_write(&F2FS_I(inode)->i_sem);
864 page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
865 if (IS_ERR(page)) {
866 err = PTR_ERR(page);
867 goto fail;
868 }
869 f2fs_put_page(page, 1);
870
871 clear_inode_flag(inode, FI_NEW_INODE);
872 f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
873 fail:
874 f2fs_up_write(&F2FS_I(inode)->i_sem);
875 return err;
876 }
877
f2fs_drop_nlink(struct inode * dir,struct inode * inode)878 void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
879 {
880 struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
881
882 f2fs_down_write(&F2FS_I(inode)->i_sem);
883
884 if (S_ISDIR(inode->i_mode))
885 f2fs_i_links_write(dir, false);
886 inode_set_ctime_current(inode);
887
888 f2fs_i_links_write(inode, false);
889 if (S_ISDIR(inode->i_mode)) {
890 f2fs_i_links_write(inode, false);
891 f2fs_i_size_write(inode, 0);
892 }
893 f2fs_up_write(&F2FS_I(inode)->i_sem);
894
895 if (inode->i_nlink == 0)
896 f2fs_add_orphan_inode(inode);
897 else
898 f2fs_release_orphan_inode(sbi);
899 }
900
901 /*
902 * It only removes the dentry from the dentry page, corresponding name
903 * entry in name page does not need to be touched during deletion.
904 */
f2fs_delete_entry(struct f2fs_dir_entry * dentry,struct page * page,struct inode * dir,struct inode * inode)905 void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
906 struct inode *dir, struct inode *inode)
907 {
908 struct f2fs_dentry_block *dentry_blk;
909 unsigned int bit_pos;
910 int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
911 int i;
912
913 f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
914
915 if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
916 f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
917
918 if (f2fs_has_inline_dentry(dir))
919 return f2fs_delete_inline_entry(dentry, page, dir, inode);
920
921 lock_page(page);
922 f2fs_wait_on_page_writeback(page, DATA, true, true);
923
924 dentry_blk = page_address(page);
925 bit_pos = dentry - dentry_blk->dentry;
926 for (i = 0; i < slots; i++)
927 __clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
928
929 /* Let's check and deallocate this dentry page */
930 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
931 NR_DENTRY_IN_BLOCK,
932 0);
933 set_page_dirty(page);
934
935 if (bit_pos == NR_DENTRY_IN_BLOCK &&
936 !f2fs_truncate_hole(dir, page->index, page->index + 1)) {
937 f2fs_clear_page_cache_dirty_tag(page);
938 clear_page_dirty_for_io(page);
939 ClearPageUptodate(page);
940 clear_page_private_all(page);
941
942 inode_dec_dirty_pages(dir);
943 f2fs_remove_dirty_inode(dir);
944 }
945 f2fs_put_page(page, 1);
946
947 dir->i_mtime = inode_set_ctime_current(dir);
948 f2fs_mark_inode_dirty_sync(dir, false);
949
950 if (inode)
951 f2fs_drop_nlink(dir, inode);
952 }
953
f2fs_empty_dir(struct inode * dir)954 bool f2fs_empty_dir(struct inode *dir)
955 {
956 unsigned long bidx = 0;
957 struct page *dentry_page;
958 unsigned int bit_pos;
959 struct f2fs_dentry_block *dentry_blk;
960 unsigned long nblock = dir_blocks(dir);
961
962 if (f2fs_has_inline_dentry(dir))
963 return f2fs_empty_inline_dir(dir);
964
965 while (bidx < nblock) {
966 pgoff_t next_pgofs;
967
968 dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
969 if (IS_ERR(dentry_page)) {
970 if (PTR_ERR(dentry_page) == -ENOENT) {
971 bidx = next_pgofs;
972 continue;
973 } else {
974 return false;
975 }
976 }
977
978 dentry_blk = page_address(dentry_page);
979 if (bidx == 0)
980 bit_pos = 2;
981 else
982 bit_pos = 0;
983 bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
984 NR_DENTRY_IN_BLOCK,
985 bit_pos);
986
987 f2fs_put_page(dentry_page, 0);
988
989 if (bit_pos < NR_DENTRY_IN_BLOCK)
990 return false;
991
992 bidx++;
993 }
994 return true;
995 }
996
f2fs_fill_dentries(struct dir_context * ctx,struct f2fs_dentry_ptr * d,unsigned int start_pos,struct fscrypt_str * fstr)997 int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
998 unsigned int start_pos, struct fscrypt_str *fstr)
999 {
1000 unsigned char d_type = DT_UNKNOWN;
1001 unsigned int bit_pos;
1002 struct f2fs_dir_entry *de = NULL;
1003 struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
1004 struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
1005 struct blk_plug plug;
1006 bool readdir_ra = sbi->readdir_ra;
1007 bool found_valid_dirent = false;
1008 int err = 0;
1009
1010 bit_pos = ((unsigned long)ctx->pos % d->max);
1011
1012 if (readdir_ra)
1013 blk_start_plug(&plug);
1014
1015 while (bit_pos < d->max) {
1016 bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
1017 if (bit_pos >= d->max)
1018 break;
1019
1020 de = &d->dentry[bit_pos];
1021 if (de->name_len == 0) {
1022 if (found_valid_dirent || !bit_pos) {
1023 f2fs_warn_ratelimited(sbi,
1024 "invalid namelen(0), ino:%u, run fsck to fix.",
1025 le32_to_cpu(de->ino));
1026 set_sbi_flag(sbi, SBI_NEED_FSCK);
1027 }
1028 bit_pos++;
1029 ctx->pos = start_pos + bit_pos;
1030 continue;
1031 }
1032
1033 d_type = fs_ftype_to_dtype(de->file_type);
1034
1035 de_name.name = d->filename[bit_pos];
1036 de_name.len = le16_to_cpu(de->name_len);
1037
1038 /* check memory boundary before moving forward */
1039 bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
1040 if (unlikely(bit_pos > d->max ||
1041 le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
1042 f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
1043 __func__, le16_to_cpu(de->name_len));
1044 set_sbi_flag(sbi, SBI_NEED_FSCK);
1045 err = -EFSCORRUPTED;
1046 f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
1047 goto out;
1048 }
1049
1050 if (IS_ENCRYPTED(d->inode)) {
1051 int save_len = fstr->len;
1052
1053 err = fscrypt_fname_disk_to_usr(d->inode,
1054 (u32)le32_to_cpu(de->hash_code),
1055 0, &de_name, fstr);
1056 if (err)
1057 goto out;
1058
1059 de_name = *fstr;
1060 fstr->len = save_len;
1061 }
1062
1063 if (!dir_emit(ctx, de_name.name, de_name.len,
1064 le32_to_cpu(de->ino), d_type)) {
1065 err = 1;
1066 goto out;
1067 }
1068
1069 if (readdir_ra)
1070 f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1071
1072 ctx->pos = start_pos + bit_pos;
1073 found_valid_dirent = true;
1074 }
1075 out:
1076 if (readdir_ra)
1077 blk_finish_plug(&plug);
1078 return err;
1079 }
1080
f2fs_readdir(struct file * file,struct dir_context * ctx)1081 static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1082 {
1083 struct inode *inode = file_inode(file);
1084 unsigned long npages = dir_blocks(inode);
1085 struct f2fs_dentry_block *dentry_blk = NULL;
1086 struct page *dentry_page = NULL;
1087 struct file_ra_state *ra = &file->f_ra;
1088 loff_t start_pos = ctx->pos;
1089 unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1090 struct f2fs_dentry_ptr d;
1091 struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1092 int err = 0;
1093
1094 if (IS_ENCRYPTED(inode)) {
1095 err = fscrypt_prepare_readdir(inode);
1096 if (err)
1097 goto out;
1098
1099 err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1100 if (err < 0)
1101 goto out;
1102 }
1103
1104 if (f2fs_has_inline_dentry(inode)) {
1105 err = f2fs_read_inline_dir(file, ctx, &fstr);
1106 goto out_free;
1107 }
1108
1109 for (; n < npages; ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1110 pgoff_t next_pgofs;
1111
1112 /* allow readdir() to be interrupted */
1113 if (fatal_signal_pending(current)) {
1114 err = -ERESTARTSYS;
1115 goto out_free;
1116 }
1117 cond_resched();
1118
1119 /* readahead for multi pages of dir */
1120 if (npages - n > 1 && !ra_has_index(ra, n))
1121 page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1122 min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1123
1124 dentry_page = f2fs_find_data_page(inode, n, &next_pgofs);
1125 if (IS_ERR(dentry_page)) {
1126 err = PTR_ERR(dentry_page);
1127 if (err == -ENOENT) {
1128 err = 0;
1129 n = next_pgofs;
1130 continue;
1131 } else {
1132 goto out_free;
1133 }
1134 }
1135
1136 dentry_blk = page_address(dentry_page);
1137
1138 make_dentry_ptr_block(inode, &d, dentry_blk);
1139
1140 err = f2fs_fill_dentries(ctx, &d,
1141 n * NR_DENTRY_IN_BLOCK, &fstr);
1142 if (err) {
1143 f2fs_put_page(dentry_page, 0);
1144 break;
1145 }
1146
1147 f2fs_put_page(dentry_page, 0);
1148
1149 n++;
1150 }
1151 out_free:
1152 fscrypt_fname_free_buffer(&fstr);
1153 out:
1154 trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1155 return err < 0 ? err : 0;
1156 }
1157
1158 const struct file_operations f2fs_dir_operations = {
1159 .llseek = generic_file_llseek,
1160 .read = generic_read_dir,
1161 .iterate_shared = f2fs_readdir,
1162 .fsync = f2fs_sync_file,
1163 .unlocked_ioctl = f2fs_ioctl,
1164 #ifdef CONFIG_COMPAT
1165 .compat_ioctl = f2fs_compat_ioctl,
1166 #endif
1167 };
1168