xref: /linux/fs/f2fs/dir.c (revision a1ff5a7d78a036d6c2178ee5acd6ba4946243800)
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 IS_ENABLED(CONFIG_UNICODE)
46 /* If @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
f2fs_init_casefolded_name(const struct inode * dir,struct f2fs_filename * 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 
f2fs_free_casefolded_name(struct f2fs_filename * fname)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 
__f2fs_setup_filename(const struct inode * dir,const struct fscrypt_name * crypt_name,struct f2fs_filename * fname)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  */
f2fs_setup_filename(struct inode * dir,const struct qstr * iname,int lookup,struct f2fs_filename * fname)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  */
f2fs_prepare_lookup(struct inode * dir,struct dentry * dentry,struct f2fs_filename * fname)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 
f2fs_free_filename(struct f2fs_filename * fname)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 
dir_block_index(unsigned int level,int dir_level,unsigned int idx)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 += dir_buckets(i, dir_level) * bucket_blocks(i);
170 	bidx += idx * bucket_blocks(level);
171 	return bidx;
172 }
173 
find_in_block(struct inode * dir,struct page * dentry_page,const struct f2fs_filename * fname,int * max_slots)174 static struct f2fs_dir_entry *find_in_block(struct inode *dir,
175 				struct page *dentry_page,
176 				const struct f2fs_filename *fname,
177 				int *max_slots)
178 {
179 	struct f2fs_dentry_block *dentry_blk;
180 	struct f2fs_dentry_ptr d;
181 
182 	dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
183 
184 	make_dentry_ptr_block(dir, &d, dentry_blk);
185 	return f2fs_find_target_dentry(&d, fname, max_slots);
186 }
187 
f2fs_match_name(const struct inode * dir,const struct f2fs_filename * fname,const u8 * de_name,u32 de_name_len)188 static inline int f2fs_match_name(const struct inode *dir,
189 				   const struct f2fs_filename *fname,
190 				   const u8 *de_name, u32 de_name_len)
191 {
192 	struct fscrypt_name f;
193 
194 #if IS_ENABLED(CONFIG_UNICODE)
195 	if (fname->cf_name.name)
196 		return generic_ci_match(dir, fname->usr_fname,
197 					&fname->cf_name,
198 					de_name, de_name_len);
199 
200 #endif
201 	f.usr_fname = fname->usr_fname;
202 	f.disk_name = fname->disk_name;
203 #ifdef CONFIG_FS_ENCRYPTION
204 	f.crypto_buf = fname->crypto_buf;
205 #endif
206 	return fscrypt_match_name(&f, de_name, de_name_len);
207 }
208 
f2fs_find_target_dentry(const struct f2fs_dentry_ptr * d,const struct f2fs_filename * fname,int * max_slots)209 struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
210 			const struct f2fs_filename *fname, int *max_slots)
211 {
212 	struct f2fs_dir_entry *de;
213 	unsigned long bit_pos = 0;
214 	int max_len = 0;
215 	int res = 0;
216 
217 	if (max_slots)
218 		*max_slots = 0;
219 	while (bit_pos < d->max) {
220 		if (!test_bit_le(bit_pos, d->bitmap)) {
221 			bit_pos++;
222 			max_len++;
223 			continue;
224 		}
225 
226 		de = &d->dentry[bit_pos];
227 
228 		if (unlikely(!de->name_len)) {
229 			bit_pos++;
230 			continue;
231 		}
232 
233 		if (de->hash_code == fname->hash) {
234 			res = f2fs_match_name(d->inode, fname,
235 					      d->filename[bit_pos],
236 					      le16_to_cpu(de->name_len));
237 			if (res < 0)
238 				return ERR_PTR(res);
239 			if (res)
240 				goto found;
241 		}
242 
243 		if (max_slots && max_len > *max_slots)
244 			*max_slots = max_len;
245 		max_len = 0;
246 
247 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
248 	}
249 
250 	de = NULL;
251 found:
252 	if (max_slots && max_len > *max_slots)
253 		*max_slots = max_len;
254 	return de;
255 }
256 
find_in_level(struct inode * dir,unsigned int level,const struct f2fs_filename * fname,struct page ** res_page)257 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
258 					unsigned int level,
259 					const struct f2fs_filename *fname,
260 					struct page **res_page)
261 {
262 	int s = GET_DENTRY_SLOTS(fname->disk_name.len);
263 	unsigned int nbucket, nblock;
264 	unsigned int bidx, end_block;
265 	struct page *dentry_page;
266 	struct f2fs_dir_entry *de = NULL;
267 	pgoff_t next_pgofs;
268 	bool room = false;
269 	int max_slots;
270 
271 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
272 	nblock = bucket_blocks(level);
273 
274 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
275 			       le32_to_cpu(fname->hash) % nbucket);
276 	end_block = bidx + nblock;
277 
278 	while (bidx < end_block) {
279 		/* no need to allocate new dentry pages to all the indices */
280 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
281 		if (IS_ERR(dentry_page)) {
282 			if (PTR_ERR(dentry_page) == -ENOENT) {
283 				room = true;
284 				bidx = next_pgofs;
285 				continue;
286 			} else {
287 				*res_page = dentry_page;
288 				break;
289 			}
290 		}
291 
292 		de = find_in_block(dir, dentry_page, fname, &max_slots);
293 		if (IS_ERR(de)) {
294 			*res_page = ERR_CAST(de);
295 			de = NULL;
296 			break;
297 		} else if (de) {
298 			*res_page = dentry_page;
299 			break;
300 		}
301 
302 		if (max_slots >= s)
303 			room = true;
304 		f2fs_put_page(dentry_page, 0);
305 
306 		bidx++;
307 	}
308 
309 	if (!de && room && F2FS_I(dir)->chash != fname->hash) {
310 		F2FS_I(dir)->chash = fname->hash;
311 		F2FS_I(dir)->clevel = level;
312 	}
313 
314 	return de;
315 }
316 
__f2fs_find_entry(struct inode * dir,const struct f2fs_filename * fname,struct page ** res_page)317 struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
318 					 const struct f2fs_filename *fname,
319 					 struct page **res_page)
320 {
321 	unsigned long npages = dir_blocks(dir);
322 	struct f2fs_dir_entry *de = NULL;
323 	unsigned int max_depth;
324 	unsigned int level;
325 
326 	*res_page = NULL;
327 
328 	if (f2fs_has_inline_dentry(dir)) {
329 		de = f2fs_find_in_inline_dir(dir, fname, res_page);
330 		goto out;
331 	}
332 
333 	if (npages == 0)
334 		goto out;
335 
336 	max_depth = F2FS_I(dir)->i_current_depth;
337 	if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
338 		f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
339 			  dir->i_ino, max_depth);
340 		max_depth = MAX_DIR_HASH_DEPTH;
341 		f2fs_i_depth_write(dir, max_depth);
342 	}
343 
344 	for (level = 0; level < max_depth; level++) {
345 		de = find_in_level(dir, level, fname, res_page);
346 		if (de || IS_ERR(*res_page))
347 			break;
348 	}
349 out:
350 	/* This is to increase the speed of f2fs_create */
351 	if (!de)
352 		F2FS_I(dir)->task = current;
353 	return de;
354 }
355 
356 /*
357  * Find an entry in the specified directory with the wanted name.
358  * It returns the page where the entry was found (as a parameter - res_page),
359  * and the entry itself. Page is returned mapped and unlocked.
360  * Entry is guaranteed to be valid.
361  */
f2fs_find_entry(struct inode * dir,const struct qstr * child,struct page ** res_page)362 struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
363 			const struct qstr *child, struct page **res_page)
364 {
365 	struct f2fs_dir_entry *de = NULL;
366 	struct f2fs_filename fname;
367 	int err;
368 
369 	err = f2fs_setup_filename(dir, child, 1, &fname);
370 	if (err) {
371 		if (err == -ENOENT)
372 			*res_page = NULL;
373 		else
374 			*res_page = ERR_PTR(err);
375 		return NULL;
376 	}
377 
378 	de = __f2fs_find_entry(dir, &fname, res_page);
379 
380 	f2fs_free_filename(&fname);
381 	return de;
382 }
383 
f2fs_parent_dir(struct inode * dir,struct page ** p)384 struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
385 {
386 	return f2fs_find_entry(dir, &dotdot_name, p);
387 }
388 
f2fs_inode_by_name(struct inode * dir,const struct qstr * qstr,struct page ** page)389 ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
390 							struct page **page)
391 {
392 	ino_t res = 0;
393 	struct f2fs_dir_entry *de;
394 
395 	de = f2fs_find_entry(dir, qstr, page);
396 	if (de) {
397 		res = le32_to_cpu(de->ino);
398 		f2fs_put_page(*page, 0);
399 	}
400 
401 	return res;
402 }
403 
f2fs_set_link(struct inode * dir,struct f2fs_dir_entry * de,struct page * page,struct inode * inode)404 void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
405 		struct page *page, struct inode *inode)
406 {
407 	enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
408 
409 	lock_page(page);
410 	f2fs_wait_on_page_writeback(page, type, true, true);
411 	de->ino = cpu_to_le32(inode->i_ino);
412 	de->file_type = fs_umode_to_ftype(inode->i_mode);
413 	set_page_dirty(page);
414 
415 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
416 	f2fs_mark_inode_dirty_sync(dir, false);
417 	f2fs_put_page(page, 1);
418 }
419 
init_dent_inode(struct inode * dir,struct inode * inode,const struct f2fs_filename * fname,struct page * ipage)420 static void init_dent_inode(struct inode *dir, struct inode *inode,
421 			    const struct f2fs_filename *fname,
422 			    struct page *ipage)
423 {
424 	struct f2fs_inode *ri;
425 
426 	if (!fname) /* tmpfile case? */
427 		return;
428 
429 	f2fs_wait_on_page_writeback(ipage, NODE, true, true);
430 
431 	/* copy name info. to this inode page */
432 	ri = F2FS_INODE(ipage);
433 	ri->i_namelen = cpu_to_le32(fname->disk_name.len);
434 	memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
435 	if (IS_ENCRYPTED(dir)) {
436 		file_set_enc_name(inode);
437 		/*
438 		 * Roll-forward recovery doesn't have encryption keys available,
439 		 * so it can't compute the dirhash for encrypted+casefolded
440 		 * filenames.  Append it to i_name if possible.  Else, disable
441 		 * roll-forward recovery of the dentry (i.e., make fsync'ing the
442 		 * file force a checkpoint) by setting LOST_PINO.
443 		 */
444 		if (IS_CASEFOLDED(dir)) {
445 			if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
446 			    F2FS_NAME_LEN)
447 				put_unaligned(fname->hash, (f2fs_hash_t *)
448 					&ri->i_name[fname->disk_name.len]);
449 			else
450 				file_lost_pino(inode);
451 		}
452 	}
453 	set_page_dirty(ipage);
454 }
455 
f2fs_do_make_empty_dir(struct inode * inode,struct inode * parent,struct f2fs_dentry_ptr * d)456 void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
457 					struct f2fs_dentry_ptr *d)
458 {
459 	struct fscrypt_str dot = FSTR_INIT(".", 1);
460 	struct fscrypt_str dotdot = FSTR_INIT("..", 2);
461 
462 	/* update dirent of "." */
463 	f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
464 
465 	/* update dirent of ".." */
466 	f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
467 }
468 
make_empty_dir(struct inode * inode,struct inode * parent,struct page * page)469 static int make_empty_dir(struct inode *inode,
470 		struct inode *parent, struct page *page)
471 {
472 	struct page *dentry_page;
473 	struct f2fs_dentry_block *dentry_blk;
474 	struct f2fs_dentry_ptr d;
475 
476 	if (f2fs_has_inline_dentry(inode))
477 		return f2fs_make_empty_inline_dir(inode, parent, page);
478 
479 	dentry_page = f2fs_get_new_data_page(inode, page, 0, true);
480 	if (IS_ERR(dentry_page))
481 		return PTR_ERR(dentry_page);
482 
483 	dentry_blk = page_address(dentry_page);
484 
485 	make_dentry_ptr_block(NULL, &d, dentry_blk);
486 	f2fs_do_make_empty_dir(inode, parent, &d);
487 
488 	set_page_dirty(dentry_page);
489 	f2fs_put_page(dentry_page, 1);
490 	return 0;
491 }
492 
f2fs_init_inode_metadata(struct inode * inode,struct inode * dir,const struct f2fs_filename * fname,struct page * dpage)493 struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
494 			const struct f2fs_filename *fname, struct page *dpage)
495 {
496 	struct page *page;
497 	int err;
498 
499 	if (is_inode_flag_set(inode, FI_NEW_INODE)) {
500 		page = f2fs_new_inode_page(inode);
501 		if (IS_ERR(page))
502 			return page;
503 
504 		if (S_ISDIR(inode->i_mode)) {
505 			/* in order to handle error case */
506 			get_page(page);
507 			err = make_empty_dir(inode, dir, page);
508 			if (err) {
509 				lock_page(page);
510 				goto put_error;
511 			}
512 			put_page(page);
513 		}
514 
515 		err = f2fs_init_acl(inode, dir, page, dpage);
516 		if (err)
517 			goto put_error;
518 
519 		err = f2fs_init_security(inode, dir,
520 					 fname ? fname->usr_fname : NULL, page);
521 		if (err)
522 			goto put_error;
523 
524 		if (IS_ENCRYPTED(inode)) {
525 			err = fscrypt_set_context(inode, page);
526 			if (err)
527 				goto put_error;
528 		}
529 	} else {
530 		page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
531 		if (IS_ERR(page))
532 			return page;
533 	}
534 
535 	init_dent_inode(dir, inode, fname, page);
536 
537 	/*
538 	 * This file should be checkpointed during fsync.
539 	 * We lost i_pino from now on.
540 	 */
541 	if (is_inode_flag_set(inode, FI_INC_LINK)) {
542 		if (!S_ISDIR(inode->i_mode))
543 			file_lost_pino(inode);
544 		/*
545 		 * If link the tmpfile to alias through linkat path,
546 		 * we should remove this inode from orphan list.
547 		 */
548 		if (inode->i_nlink == 0)
549 			f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
550 		f2fs_i_links_write(inode, true);
551 	}
552 	return page;
553 
554 put_error:
555 	clear_nlink(inode);
556 	f2fs_update_inode(inode, page);
557 	f2fs_put_page(page, 1);
558 	return ERR_PTR(err);
559 }
560 
f2fs_update_parent_metadata(struct inode * dir,struct inode * inode,unsigned int current_depth)561 void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
562 						unsigned int current_depth)
563 {
564 	if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
565 		if (S_ISDIR(inode->i_mode))
566 			f2fs_i_links_write(dir, true);
567 		clear_inode_flag(inode, FI_NEW_INODE);
568 	}
569 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
570 	f2fs_mark_inode_dirty_sync(dir, false);
571 
572 	if (F2FS_I(dir)->i_current_depth != current_depth)
573 		f2fs_i_depth_write(dir, current_depth);
574 
575 	if (inode && is_inode_flag_set(inode, FI_INC_LINK))
576 		clear_inode_flag(inode, FI_INC_LINK);
577 }
578 
f2fs_room_for_filename(const void * bitmap,int slots,int max_slots)579 int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
580 {
581 	int bit_start = 0;
582 	int zero_start, zero_end;
583 next:
584 	zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
585 	if (zero_start >= max_slots)
586 		return max_slots;
587 
588 	zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
589 	if (zero_end - zero_start >= slots)
590 		return zero_start;
591 
592 	bit_start = zero_end + 1;
593 
594 	if (zero_end + 1 >= max_slots)
595 		return max_slots;
596 	goto next;
597 }
598 
f2fs_has_enough_room(struct inode * dir,struct page * ipage,const struct f2fs_filename * fname)599 bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
600 			  const struct f2fs_filename *fname)
601 {
602 	struct f2fs_dentry_ptr d;
603 	unsigned int bit_pos;
604 	int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
605 
606 	make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
607 
608 	bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
609 
610 	return bit_pos < d.max;
611 }
612 
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)613 void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
614 			const struct fscrypt_str *name, f2fs_hash_t name_hash,
615 			unsigned int bit_pos)
616 {
617 	struct f2fs_dir_entry *de;
618 	int slots = GET_DENTRY_SLOTS(name->len);
619 	int i;
620 
621 	de = &d->dentry[bit_pos];
622 	de->hash_code = name_hash;
623 	de->name_len = cpu_to_le16(name->len);
624 	memcpy(d->filename[bit_pos], name->name, name->len);
625 	de->ino = cpu_to_le32(ino);
626 	de->file_type = fs_umode_to_ftype(mode);
627 	for (i = 0; i < slots; i++) {
628 		__set_bit_le(bit_pos + i, (void *)d->bitmap);
629 		/* avoid wrong garbage data for readdir */
630 		if (i)
631 			(de + i)->name_len = 0;
632 	}
633 }
634 
f2fs_add_regular_entry(struct inode * dir,const struct f2fs_filename * fname,struct inode * inode,nid_t ino,umode_t mode)635 int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
636 			   struct inode *inode, nid_t ino, umode_t mode)
637 {
638 	unsigned int bit_pos;
639 	unsigned int level;
640 	unsigned int current_depth;
641 	unsigned long bidx, block;
642 	unsigned int nbucket, nblock;
643 	struct page *dentry_page = NULL;
644 	struct f2fs_dentry_block *dentry_blk = NULL;
645 	struct f2fs_dentry_ptr d;
646 	struct page *page = NULL;
647 	int slots, err = 0;
648 
649 	level = 0;
650 	slots = GET_DENTRY_SLOTS(fname->disk_name.len);
651 
652 	current_depth = F2FS_I(dir)->i_current_depth;
653 	if (F2FS_I(dir)->chash == fname->hash) {
654 		level = F2FS_I(dir)->clevel;
655 		F2FS_I(dir)->chash = 0;
656 	}
657 
658 start:
659 	if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH))
660 		return -ENOSPC;
661 
662 	if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
663 		return -ENOSPC;
664 
665 	/* Increase the depth, if required */
666 	if (level == current_depth)
667 		++current_depth;
668 
669 	nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
670 	nblock = bucket_blocks(level);
671 
672 	bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
673 				(le32_to_cpu(fname->hash) % nbucket));
674 
675 	for (block = bidx; block <= (bidx + nblock - 1); block++) {
676 		dentry_page = f2fs_get_new_data_page(dir, NULL, block, true);
677 		if (IS_ERR(dentry_page))
678 			return PTR_ERR(dentry_page);
679 
680 		dentry_blk = page_address(dentry_page);
681 		bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
682 						slots, NR_DENTRY_IN_BLOCK);
683 		if (bit_pos < NR_DENTRY_IN_BLOCK)
684 			goto add_dentry;
685 
686 		f2fs_put_page(dentry_page, 1);
687 	}
688 
689 	/* Move to next level to find the empty slot for new dentry */
690 	++level;
691 	goto start;
692 add_dentry:
693 	f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
694 
695 	if (inode) {
696 		f2fs_down_write(&F2FS_I(inode)->i_sem);
697 		page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
698 		if (IS_ERR(page)) {
699 			err = PTR_ERR(page);
700 			goto fail;
701 		}
702 	}
703 
704 	make_dentry_ptr_block(NULL, &d, dentry_blk);
705 	f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
706 			   bit_pos);
707 
708 	set_page_dirty(dentry_page);
709 
710 	if (inode) {
711 		f2fs_i_pino_write(inode, dir->i_ino);
712 
713 		/* synchronize inode page's data from inode cache */
714 		if (is_inode_flag_set(inode, FI_NEW_INODE))
715 			f2fs_update_inode(inode, page);
716 
717 		f2fs_put_page(page, 1);
718 	}
719 
720 	f2fs_update_parent_metadata(dir, inode, current_depth);
721 fail:
722 	if (inode)
723 		f2fs_up_write(&F2FS_I(inode)->i_sem);
724 
725 	f2fs_put_page(dentry_page, 1);
726 
727 	return err;
728 }
729 
f2fs_add_dentry(struct inode * dir,const struct f2fs_filename * fname,struct inode * inode,nid_t ino,umode_t mode)730 int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
731 		    struct inode *inode, nid_t ino, umode_t mode)
732 {
733 	int err = -EAGAIN;
734 
735 	if (f2fs_has_inline_dentry(dir)) {
736 		/*
737 		 * Should get i_xattr_sem to keep the lock order:
738 		 * i_xattr_sem -> inode_page lock used by f2fs_setxattr.
739 		 */
740 		f2fs_down_read(&F2FS_I(dir)->i_xattr_sem);
741 		err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
742 		f2fs_up_read(&F2FS_I(dir)->i_xattr_sem);
743 	}
744 	if (err == -EAGAIN)
745 		err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
746 
747 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
748 	return err;
749 }
750 
751 /*
752  * Caller should grab and release a rwsem by calling f2fs_lock_op() and
753  * f2fs_unlock_op().
754  */
f2fs_do_add_link(struct inode * dir,const struct qstr * name,struct inode * inode,nid_t ino,umode_t mode)755 int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
756 				struct inode *inode, nid_t ino, umode_t mode)
757 {
758 	struct f2fs_filename fname;
759 	struct page *page = NULL;
760 	struct f2fs_dir_entry *de = NULL;
761 	int err;
762 
763 	err = f2fs_setup_filename(dir, name, 0, &fname);
764 	if (err)
765 		return err;
766 
767 	/*
768 	 * An immature stackable filesystem shows a race condition between lookup
769 	 * and create. If we have same task when doing lookup and create, it's
770 	 * definitely fine as expected by VFS normally. Otherwise, let's just
771 	 * verify on-disk dentry one more time, which guarantees filesystem
772 	 * consistency more.
773 	 */
774 	if (current != F2FS_I(dir)->task) {
775 		de = __f2fs_find_entry(dir, &fname, &page);
776 		F2FS_I(dir)->task = NULL;
777 	}
778 	if (de) {
779 		f2fs_put_page(page, 0);
780 		err = -EEXIST;
781 	} else if (IS_ERR(page)) {
782 		err = PTR_ERR(page);
783 	} else {
784 		err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
785 	}
786 	f2fs_free_filename(&fname);
787 	return err;
788 }
789 
f2fs_do_tmpfile(struct inode * inode,struct inode * dir,struct f2fs_filename * fname)790 int f2fs_do_tmpfile(struct inode *inode, struct inode *dir,
791 					struct f2fs_filename *fname)
792 {
793 	struct page *page;
794 	int err = 0;
795 
796 	f2fs_down_write(&F2FS_I(inode)->i_sem);
797 	page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
798 	if (IS_ERR(page)) {
799 		err = PTR_ERR(page);
800 		goto fail;
801 	}
802 	f2fs_put_page(page, 1);
803 
804 	clear_inode_flag(inode, FI_NEW_INODE);
805 	f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
806 fail:
807 	f2fs_up_write(&F2FS_I(inode)->i_sem);
808 	return err;
809 }
810 
f2fs_drop_nlink(struct inode * dir,struct inode * inode)811 void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
812 {
813 	struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
814 
815 	f2fs_down_write(&F2FS_I(inode)->i_sem);
816 
817 	if (S_ISDIR(inode->i_mode))
818 		f2fs_i_links_write(dir, false);
819 	inode_set_ctime_current(inode);
820 
821 	f2fs_i_links_write(inode, false);
822 	if (S_ISDIR(inode->i_mode)) {
823 		f2fs_i_links_write(inode, false);
824 		f2fs_i_size_write(inode, 0);
825 	}
826 	f2fs_up_write(&F2FS_I(inode)->i_sem);
827 
828 	if (inode->i_nlink == 0)
829 		f2fs_add_orphan_inode(inode);
830 	else
831 		f2fs_release_orphan_inode(sbi);
832 }
833 
834 /*
835  * It only removes the dentry from the dentry page, corresponding name
836  * entry in name page does not need to be touched during deletion.
837  */
f2fs_delete_entry(struct f2fs_dir_entry * dentry,struct page * page,struct inode * dir,struct inode * inode)838 void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
839 					struct inode *dir, struct inode *inode)
840 {
841 	struct	f2fs_dentry_block *dentry_blk;
842 	unsigned int bit_pos;
843 	int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
844 	int i;
845 
846 	f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
847 
848 	if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
849 		f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
850 
851 	if (f2fs_has_inline_dentry(dir))
852 		return f2fs_delete_inline_entry(dentry, page, dir, inode);
853 
854 	lock_page(page);
855 	f2fs_wait_on_page_writeback(page, DATA, true, true);
856 
857 	dentry_blk = page_address(page);
858 	bit_pos = dentry - dentry_blk->dentry;
859 	for (i = 0; i < slots; i++)
860 		__clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
861 
862 	/* Let's check and deallocate this dentry page */
863 	bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
864 			NR_DENTRY_IN_BLOCK,
865 			0);
866 	set_page_dirty(page);
867 
868 	if (bit_pos == NR_DENTRY_IN_BLOCK &&
869 		!f2fs_truncate_hole(dir, page->index, page->index + 1)) {
870 		f2fs_clear_page_cache_dirty_tag(page);
871 		clear_page_dirty_for_io(page);
872 		ClearPageUptodate(page);
873 		clear_page_private_all(page);
874 
875 		inode_dec_dirty_pages(dir);
876 		f2fs_remove_dirty_inode(dir);
877 	}
878 	f2fs_put_page(page, 1);
879 
880 	inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
881 	f2fs_mark_inode_dirty_sync(dir, false);
882 
883 	if (inode)
884 		f2fs_drop_nlink(dir, inode);
885 }
886 
f2fs_empty_dir(struct inode * dir)887 bool f2fs_empty_dir(struct inode *dir)
888 {
889 	unsigned long bidx = 0;
890 	struct page *dentry_page;
891 	unsigned int bit_pos;
892 	struct f2fs_dentry_block *dentry_blk;
893 	unsigned long nblock = dir_blocks(dir);
894 
895 	if (f2fs_has_inline_dentry(dir))
896 		return f2fs_empty_inline_dir(dir);
897 
898 	while (bidx < nblock) {
899 		pgoff_t next_pgofs;
900 
901 		dentry_page = f2fs_find_data_page(dir, bidx, &next_pgofs);
902 		if (IS_ERR(dentry_page)) {
903 			if (PTR_ERR(dentry_page) == -ENOENT) {
904 				bidx = next_pgofs;
905 				continue;
906 			} else {
907 				return false;
908 			}
909 		}
910 
911 		dentry_blk = page_address(dentry_page);
912 		if (bidx == 0)
913 			bit_pos = 2;
914 		else
915 			bit_pos = 0;
916 		bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
917 						NR_DENTRY_IN_BLOCK,
918 						bit_pos);
919 
920 		f2fs_put_page(dentry_page, 0);
921 
922 		if (bit_pos < NR_DENTRY_IN_BLOCK)
923 			return false;
924 
925 		bidx++;
926 	}
927 	return true;
928 }
929 
f2fs_fill_dentries(struct dir_context * ctx,struct f2fs_dentry_ptr * d,unsigned int start_pos,struct fscrypt_str * fstr)930 int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
931 			unsigned int start_pos, struct fscrypt_str *fstr)
932 {
933 	unsigned char d_type = DT_UNKNOWN;
934 	unsigned int bit_pos;
935 	struct f2fs_dir_entry *de = NULL;
936 	struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
937 	struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
938 	struct blk_plug plug;
939 	bool readdir_ra = sbi->readdir_ra;
940 	bool found_valid_dirent = false;
941 	int err = 0;
942 
943 	bit_pos = ((unsigned long)ctx->pos % d->max);
944 
945 	if (readdir_ra)
946 		blk_start_plug(&plug);
947 
948 	while (bit_pos < d->max) {
949 		bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
950 		if (bit_pos >= d->max)
951 			break;
952 
953 		de = &d->dentry[bit_pos];
954 		if (de->name_len == 0) {
955 			if (found_valid_dirent || !bit_pos) {
956 				f2fs_warn_ratelimited(sbi,
957 					"invalid namelen(0), ino:%u, run fsck to fix.",
958 					le32_to_cpu(de->ino));
959 				set_sbi_flag(sbi, SBI_NEED_FSCK);
960 			}
961 			bit_pos++;
962 			ctx->pos = start_pos + bit_pos;
963 			continue;
964 		}
965 
966 		d_type = fs_ftype_to_dtype(de->file_type);
967 
968 		de_name.name = d->filename[bit_pos];
969 		de_name.len = le16_to_cpu(de->name_len);
970 
971 		/* check memory boundary before moving forward */
972 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
973 		if (unlikely(bit_pos > d->max ||
974 				le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
975 			f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
976 				  __func__, le16_to_cpu(de->name_len));
977 			set_sbi_flag(sbi, SBI_NEED_FSCK);
978 			err = -EFSCORRUPTED;
979 			f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
980 			goto out;
981 		}
982 
983 		if (IS_ENCRYPTED(d->inode)) {
984 			int save_len = fstr->len;
985 
986 			err = fscrypt_fname_disk_to_usr(d->inode,
987 						(u32)le32_to_cpu(de->hash_code),
988 						0, &de_name, fstr);
989 			if (err)
990 				goto out;
991 
992 			de_name = *fstr;
993 			fstr->len = save_len;
994 		}
995 
996 		if (!dir_emit(ctx, de_name.name, de_name.len,
997 					le32_to_cpu(de->ino), d_type)) {
998 			err = 1;
999 			goto out;
1000 		}
1001 
1002 		if (readdir_ra)
1003 			f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
1004 
1005 		ctx->pos = start_pos + bit_pos;
1006 		found_valid_dirent = true;
1007 	}
1008 out:
1009 	if (readdir_ra)
1010 		blk_finish_plug(&plug);
1011 	return err;
1012 }
1013 
f2fs_readdir(struct file * file,struct dir_context * ctx)1014 static int f2fs_readdir(struct file *file, struct dir_context *ctx)
1015 {
1016 	struct inode *inode = file_inode(file);
1017 	unsigned long npages = dir_blocks(inode);
1018 	struct f2fs_dentry_block *dentry_blk = NULL;
1019 	struct page *dentry_page = NULL;
1020 	struct file_ra_state *ra = &file->f_ra;
1021 	loff_t start_pos = ctx->pos;
1022 	unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
1023 	struct f2fs_dentry_ptr d;
1024 	struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
1025 	int err = 0;
1026 
1027 	if (IS_ENCRYPTED(inode)) {
1028 		err = fscrypt_prepare_readdir(inode);
1029 		if (err)
1030 			goto out;
1031 
1032 		err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
1033 		if (err < 0)
1034 			goto out;
1035 	}
1036 
1037 	if (f2fs_has_inline_dentry(inode)) {
1038 		err = f2fs_read_inline_dir(file, ctx, &fstr);
1039 		goto out_free;
1040 	}
1041 
1042 	for (; n < npages; ctx->pos = n * NR_DENTRY_IN_BLOCK) {
1043 		pgoff_t next_pgofs;
1044 
1045 		/* allow readdir() to be interrupted */
1046 		if (fatal_signal_pending(current)) {
1047 			err = -ERESTARTSYS;
1048 			goto out_free;
1049 		}
1050 		cond_resched();
1051 
1052 		/* readahead for multi pages of dir */
1053 		if (npages - n > 1 && !ra_has_index(ra, n))
1054 			page_cache_sync_readahead(inode->i_mapping, ra, file, n,
1055 				min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
1056 
1057 		dentry_page = f2fs_find_data_page(inode, n, &next_pgofs);
1058 		if (IS_ERR(dentry_page)) {
1059 			err = PTR_ERR(dentry_page);
1060 			if (err == -ENOENT) {
1061 				err = 0;
1062 				n = next_pgofs;
1063 				continue;
1064 			} else {
1065 				goto out_free;
1066 			}
1067 		}
1068 
1069 		dentry_blk = page_address(dentry_page);
1070 
1071 		make_dentry_ptr_block(inode, &d, dentry_blk);
1072 
1073 		err = f2fs_fill_dentries(ctx, &d,
1074 				n * NR_DENTRY_IN_BLOCK, &fstr);
1075 		if (err) {
1076 			f2fs_put_page(dentry_page, 0);
1077 			break;
1078 		}
1079 
1080 		f2fs_put_page(dentry_page, 0);
1081 
1082 		n++;
1083 	}
1084 out_free:
1085 	fscrypt_fname_free_buffer(&fstr);
1086 out:
1087 	trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
1088 	return err < 0 ? err : 0;
1089 }
1090 
1091 const struct file_operations f2fs_dir_operations = {
1092 	.llseek		= generic_file_llseek,
1093 	.read		= generic_read_dir,
1094 	.iterate_shared	= f2fs_readdir,
1095 	.fsync		= f2fs_sync_file,
1096 	.unlocked_ioctl	= f2fs_ioctl,
1097 #ifdef CONFIG_COMPAT
1098 	.compat_ioctl   = f2fs_compat_ioctl,
1099 #endif
1100 };
1101