projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: update f2fs document
[linux-2.6-block.git] / fs / f2fs / node.c
CommitLineData
0a8165d7 1/*
e05df3b1
JK		 2 * fs/f2fs/node.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 <linux/mpage.h>
14#include <linux/backing-dev.h>
15#include <linux/blkdev.h>
16#include <linux/pagevec.h>
17#include <linux/swap.h>
18
19#include "f2fs.h"
20#include "node.h"
21#include "segment.h"
51dd6249 22#include <trace/events/f2fs.h>
e05df3b1
JK		 23
24static struct kmem_cache *nat_entry_slab;
25static struct kmem_cache *free_nid_slab;
26
27static void clear_node_page_dirty(struct page *page)
28{
29 struct address_space *mapping = page->mapping;
30 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
31 unsigned int long flags;
32
33 if (PageDirty(page)) {
34 spin_lock_irqsave(&mapping->tree_lock, flags);
35 radix_tree_tag_clear(&mapping->page_tree,
36 page_index(page),
37 PAGECACHE_TAG_DIRTY);
38 spin_unlock_irqrestore(&mapping->tree_lock, flags);
39
40 clear_page_dirty_for_io(page);
41 dec_page_count(sbi, F2FS_DIRTY_NODES);
42 }
43 ClearPageUptodate(page);
44}
45
46static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
47{
48 pgoff_t index = current_nat_addr(sbi, nid);
49 return get_meta_page(sbi, index);
50}
51
52static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
53{
54 struct page *src_page;
55 struct page *dst_page;
56 pgoff_t src_off;
57 pgoff_t dst_off;
58 void *src_addr;
59 void *dst_addr;
60 struct f2fs_nm_info *nm_i = NM_I(sbi);
61
62 src_off = current_nat_addr(sbi, nid);
63 dst_off = next_nat_addr(sbi, src_off);
64
65 /* get current nat block page with lock */
66 src_page = get_meta_page(sbi, src_off);
67
68 /* Dirty src_page means that it is already the new target NAT page. */
69 if (PageDirty(src_page))
70 return src_page;
71
72 dst_page = grab_meta_page(sbi, dst_off);
73
74 src_addr = page_address(src_page);
75 dst_addr = page_address(dst_page);
76 memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE);
77 set_page_dirty(dst_page);
78 f2fs_put_page(src_page, 1);
79
80 set_to_next_nat(nm_i, nid);
81
82 return dst_page;
83}
84
0a8165d7 85/*
e05df3b1
JK		 86 * Readahead NAT pages
87 */
88static void ra_nat_pages(struct f2fs_sb_info *sbi, int nid)
89{
90 struct address_space *mapping = sbi->meta_inode->i_mapping;
91 struct f2fs_nm_info *nm_i = NM_I(sbi);
c718379b 92 struct blk_plug plug;
e05df3b1
JK		 93 struct page *page;
94 pgoff_t index;
95 int i;
96
c718379b
JK		 97 blk_start_plug(&plug);
98
e05df3b1
JK		 99 for (i = 0; i < FREE_NID_PAGES; i++, nid += NAT_ENTRY_PER_BLOCK) {
100 if (nid >= nm_i->max_nid)
101 nid = 0;
102 index = current_nat_addr(sbi, nid);
103
104 page = grab_cache_page(mapping, index);
105 if (!page)
106 continue;
393ff91f 107 if (PageUptodate(page)) {
e05df3b1
JK		 108 f2fs_put_page(page, 1);
109 continue;
110 }
393ff91f
JK		 111 if (f2fs_readpage(sbi, page, index, READ))
112 continue;
113
369a708c 114 f2fs_put_page(page, 0);
e05df3b1 115 }
c718379b 116 blk_finish_plug(&plug);
e05df3b1
JK		 117}
118
119static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
120{
121 return radix_tree_lookup(&nm_i->nat_root, n);
122}
123
124static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
125 nid_t start, unsigned int nr, struct nat_entry **ep)
126{
127 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
128}
129
130static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
131{
132 list_del(&e->list);
133 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
134 nm_i->nat_cnt--;
135 kmem_cache_free(nat_entry_slab, e);
136}
137
138int is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
139{
140 struct f2fs_nm_info *nm_i = NM_I(sbi);
141 struct nat_entry *e;
142 int is_cp = 1;
143
144 read_lock(&nm_i->nat_tree_lock);
145 e = __lookup_nat_cache(nm_i, nid);
146 if (e && !e->checkpointed)
147 is_cp = 0;
148 read_unlock(&nm_i->nat_tree_lock);
149 return is_cp;
150}
151
152static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid)
153{
154 struct nat_entry *new;
155
156 new = kmem_cache_alloc(nat_entry_slab, GFP_ATOMIC);
157 if (!new)
158 return NULL;
159 if (radix_tree_insert(&nm_i->nat_root, nid, new)) {
160 kmem_cache_free(nat_entry_slab, new);
161 return NULL;
162 }
163 memset(new, 0, sizeof(struct nat_entry));
164 nat_set_nid(new, nid);
165 list_add_tail(&new->list, &nm_i->nat_entries);
166 nm_i->nat_cnt++;
167 return new;
168}
169
170static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid,
171 struct f2fs_nat_entry *ne)
172{
173 struct nat_entry *e;
174retry:
175 write_lock(&nm_i->nat_tree_lock);
176 e = __lookup_nat_cache(nm_i, nid);
177 if (!e) {
178 e = grab_nat_entry(nm_i, nid);
179 if (!e) {
180 write_unlock(&nm_i->nat_tree_lock);
181 goto retry;
182 }
183 nat_set_blkaddr(e, le32_to_cpu(ne->block_addr));
184 nat_set_ino(e, le32_to_cpu(ne->ino));
185 nat_set_version(e, ne->version);
186 e->checkpointed = true;
187 }
188 write_unlock(&nm_i->nat_tree_lock);
189}
190
191static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
192 block_t new_blkaddr)
193{
194 struct f2fs_nm_info *nm_i = NM_I(sbi);
195 struct nat_entry *e;
196retry:
197 write_lock(&nm_i->nat_tree_lock);
198 e = __lookup_nat_cache(nm_i, ni->nid);
199 if (!e) {
200 e = grab_nat_entry(nm_i, ni->nid);
201 if (!e) {
202 write_unlock(&nm_i->nat_tree_lock);
203 goto retry;
204 }
205 e->ni = *ni;
206 e->checkpointed = true;
5d56b671 207 f2fs_bug_on(ni->blk_addr == NEW_ADDR);
e05df3b1
JK		 208 } else if (new_blkaddr == NEW_ADDR) {
209 /*
210 * when nid is reallocated,
211 * previous nat entry can be remained in nat cache.
212 * So, reinitialize it with new information.
213 */
214 e->ni = *ni;
5d56b671 215 f2fs_bug_on(ni->blk_addr != NULL_ADDR);
e05df3b1
JK		 216 }
217
218 if (new_blkaddr == NEW_ADDR)
219 e->checkpointed = false;
220
221 /* sanity check */
5d56b671
JK		 222 f2fs_bug_on(nat_get_blkaddr(e) != ni->blk_addr);
223 f2fs_bug_on(nat_get_blkaddr(e) == NULL_ADDR &&
e05df3b1 224 new_blkaddr == NULL_ADDR);
5d56b671 225 f2fs_bug_on(nat_get_blkaddr(e) == NEW_ADDR &&
e05df3b1 226 new_blkaddr == NEW_ADDR);
5d56b671 227 f2fs_bug_on(nat_get_blkaddr(e) != NEW_ADDR &&
e05df3b1
JK		 228 nat_get_blkaddr(e) != NULL_ADDR &&
229 new_blkaddr == NEW_ADDR);
230
231 /* increament version no as node is removed */
232 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
233 unsigned char version = nat_get_version(e);
234 nat_set_version(e, inc_node_version(version));
235 }
236
237 /* change address */
238 nat_set_blkaddr(e, new_blkaddr);
239 __set_nat_cache_dirty(nm_i, e);
240 write_unlock(&nm_i->nat_tree_lock);
241}
242
4660f9c0 243int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
e05df3b1
JK		 244{
245 struct f2fs_nm_info *nm_i = NM_I(sbi);
246
6cac3759 247 if (nm_i->nat_cnt <= NM_WOUT_THRESHOLD)
e05df3b1
JK		 248 return 0;
249
250 write_lock(&nm_i->nat_tree_lock);
251 while (nr_shrink && !list_empty(&nm_i->nat_entries)) {
252 struct nat_entry *ne;
253 ne = list_first_entry(&nm_i->nat_entries,
254 struct nat_entry, list);
255 __del_from_nat_cache(nm_i, ne);
256 nr_shrink--;
257 }
258 write_unlock(&nm_i->nat_tree_lock);
259 return nr_shrink;
260}
261
0a8165d7 262/*
e05df3b1
JK		 263 * This function returns always success
264 */
265void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
266{
267 struct f2fs_nm_info *nm_i = NM_I(sbi);
268 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
269 struct f2fs_summary_block *sum = curseg->sum_blk;
270 nid_t start_nid = START_NID(nid);
271 struct f2fs_nat_block *nat_blk;
272 struct page *page = NULL;
273 struct f2fs_nat_entry ne;
274 struct nat_entry *e;
275 int i;
276
be4124f8 277 memset(&ne, 0, sizeof(struct f2fs_nat_entry));
e05df3b1
JK		 278 ni->nid = nid;
279
280 /* Check nat cache */
281 read_lock(&nm_i->nat_tree_lock);
282 e = __lookup_nat_cache(nm_i, nid);
283 if (e) {
284 ni->ino = nat_get_ino(e);
285 ni->blk_addr = nat_get_blkaddr(e);
286 ni->version = nat_get_version(e);
287 }
288 read_unlock(&nm_i->nat_tree_lock);
289 if (e)
290 return;
291
292 /* Check current segment summary */
293 mutex_lock(&curseg->curseg_mutex);
294 i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0);
295 if (i >= 0) {
296 ne = nat_in_journal(sum, i);
297 node_info_from_raw_nat(ni, &ne);
298 }
299 mutex_unlock(&curseg->curseg_mutex);
300 if (i >= 0)
301 goto cache;
302
303 /* Fill node_info from nat page */
304 page = get_current_nat_page(sbi, start_nid);
305 nat_blk = (struct f2fs_nat_block *)page_address(page);
306 ne = nat_blk->entries[nid - start_nid];
307 node_info_from_raw_nat(ni, &ne);
308 f2fs_put_page(page, 1);
309cache:
310 /* cache nat entry */
311 cache_nat_entry(NM_I(sbi), nid, &ne);
312}
313
0a8165d7 314/*
e05df3b1
JK		 315 * The maximum depth is four.
316 * Offset[0] will have raw inode offset.
317 */
de93653f
JK		 318static int get_node_path(struct f2fs_inode_info *fi, long block,
319 int offset[4], unsigned int noffset[4])
e05df3b1 320{
de93653f 321 const long direct_index = ADDRS_PER_INODE(fi);
e05df3b1
JK		 322 const long direct_blks = ADDRS_PER_BLOCK;
323 const long dptrs_per_blk = NIDS_PER_BLOCK;
324 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
325 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
326 int n = 0;
327 int level = 0;
328
329 noffset[0] = 0;
330
331 if (block < direct_index) {
25c0a6e5 332 offset[n] = block;
e05df3b1
JK		 333 goto got;
334 }
335 block -= direct_index;
336 if (block < direct_blks) {
337 offset[n++] = NODE_DIR1_BLOCK;
338 noffset[n] = 1;
25c0a6e5 339 offset[n] = block;
e05df3b1
JK		 340 level = 1;
341 goto got;
342 }
343 block -= direct_blks;
344 if (block < direct_blks) {
345 offset[n++] = NODE_DIR2_BLOCK;
346 noffset[n] = 2;
25c0a6e5 347 offset[n] = block;
e05df3b1
JK		 348 level = 1;
349 goto got;
350 }
351 block -= direct_blks;
352 if (block < indirect_blks) {
353 offset[n++] = NODE_IND1_BLOCK;
354 noffset[n] = 3;
355 offset[n++] = block / direct_blks;
356 noffset[n] = 4 + offset[n - 1];
25c0a6e5 357 offset[n] = block % direct_blks;
e05df3b1
JK		 358 level = 2;
359 goto got;
360 }
361 block -= indirect_blks;
362 if (block < indirect_blks) {
363 offset[n++] = NODE_IND2_BLOCK;
364 noffset[n] = 4 + dptrs_per_blk;
365 offset[n++] = block / direct_blks;
366 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
25c0a6e5 367 offset[n] = block % direct_blks;
e05df3b1
JK		 368 level = 2;
369 goto got;
370 }
371 block -= indirect_blks;
372 if (block < dindirect_blks) {
373 offset[n++] = NODE_DIND_BLOCK;
374 noffset[n] = 5 + (dptrs_per_blk * 2);
375 offset[n++] = block / indirect_blks;
376 noffset[n] = 6 + (dptrs_per_blk * 2) +
377 offset[n - 1] * (dptrs_per_blk + 1);
378 offset[n++] = (block / direct_blks) % dptrs_per_blk;
379 noffset[n] = 7 + (dptrs_per_blk * 2) +
380 offset[n - 2] * (dptrs_per_blk + 1) +
381 offset[n - 1];
25c0a6e5 382 offset[n] = block % direct_blks;
e05df3b1
JK		 383 level = 3;
384 goto got;
385 } else {
386 BUG();
387 }
388got:
389 return level;
390}
391
392/*
393 * Caller should call f2fs_put_dnode(dn).
39936837
JK		 394 * Also, it should grab and release a mutex by calling mutex_lock_op() and
395 * mutex_unlock_op() only if ro is not set RDONLY_NODE.
396 * In the case of RDONLY_NODE, we don't need to care about mutex.
e05df3b1 397 */
266e97a8 398int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
e05df3b1
JK		 399{
400 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
401 struct page *npage[4];
402 struct page *parent;
403 int offset[4];
404 unsigned int noffset[4];
405 nid_t nids[4];
406 int level, i;
407 int err = 0;
408
de93653f 409 level = get_node_path(F2FS_I(dn->inode), index, offset, noffset);
e05df3b1
JK		 410
411 nids[0] = dn->inode->i_ino;
1646cfac 412 npage[0] = dn->inode_page;
e05df3b1 413
1646cfac
JK		 414 if (!npage[0]) {
415 npage[0] = get_node_page(sbi, nids[0]);
416 if (IS_ERR(npage[0]))
417 return PTR_ERR(npage[0]);
418 }
e05df3b1 419 parent = npage[0];
52c2db3f
CL		 420 if (level != 0)
421 nids[1] = get_nid(parent, offset[0], true);
e05df3b1
JK		 422 dn->inode_page = npage[0];
423 dn->inode_page_locked = true;
424
425 /* get indirect or direct nodes */
426 for (i = 1; i <= level; i++) {
427 bool done = false;
428
266e97a8 429 if (!nids[i] && mode == ALLOC_NODE) {
e05df3b1
JK		 430 /* alloc new node */
431 if (!alloc_nid(sbi, &(nids[i]))) {
e05df3b1
JK		 432 err = -ENOSPC;
433 goto release_pages;
434 }
435
436 dn->nid = nids[i];
8ae8f162 437 npage[i] = new_node_page(dn, noffset[i], NULL);
e05df3b1
JK		 438 if (IS_ERR(npage[i])) {
439 alloc_nid_failed(sbi, nids[i]);
e05df3b1
JK		 440 err = PTR_ERR(npage[i]);
441 goto release_pages;
442 }
443
444 set_nid(parent, offset[i - 1], nids[i], i == 1);
445 alloc_nid_done(sbi, nids[i]);
e05df3b1 446 done = true;
266e97a8 447 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
e05df3b1
JK		 448 npage[i] = get_node_page_ra(parent, offset[i - 1]);
449 if (IS_ERR(npage[i])) {
450 err = PTR_ERR(npage[i]);
451 goto release_pages;
452 }
453 done = true;
454 }
455 if (i == 1) {
456 dn->inode_page_locked = false;
457 unlock_page(parent);
458 } else {
459 f2fs_put_page(parent, 1);
460 }
461
462 if (!done) {
463 npage[i] = get_node_page(sbi, nids[i]);
464 if (IS_ERR(npage[i])) {
465 err = PTR_ERR(npage[i]);
466 f2fs_put_page(npage[0], 0);
467 goto release_out;
468 }
469 }
470 if (i < level) {
471 parent = npage[i];
472 nids[i + 1] = get_nid(parent, offset[i], false);
473 }
474 }
475 dn->nid = nids[level];
476 dn->ofs_in_node = offset[level];
477 dn->node_page = npage[level];
478 dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node);
479 return 0;
480
481release_pages:
482 f2fs_put_page(parent, 1);
483 if (i > 1)
484 f2fs_put_page(npage[0], 0);
485release_out:
486 dn->inode_page = NULL;
487 dn->node_page = NULL;
488 return err;
489}
490
491static void truncate_node(struct dnode_of_data *dn)
492{
493 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
494 struct node_info ni;
495
496 get_node_info(sbi, dn->nid, &ni);
71e9fec5 497 if (dn->inode->i_blocks == 0) {
5d56b671 498 f2fs_bug_on(ni.blk_addr != NULL_ADDR);
71e9fec5
JK		 499 goto invalidate;
500 }
5d56b671 501 f2fs_bug_on(ni.blk_addr == NULL_ADDR);
e05df3b1 502
e05df3b1 503 /* Deallocate node address */
71e9fec5 504 invalidate_blocks(sbi, ni.blk_addr);
e05df3b1
JK		 505 dec_valid_node_count(sbi, dn->inode, 1);
506 set_node_addr(sbi, &ni, NULL_ADDR);
507
508 if (dn->nid == dn->inode->i_ino) {
509 remove_orphan_inode(sbi, dn->nid);
510 dec_valid_inode_count(sbi);
511 } else {
512 sync_inode_page(dn);
513 }
71e9fec5 514invalidate:
e05df3b1
JK		 515 clear_node_page_dirty(dn->node_page);
516 F2FS_SET_SB_DIRT(sbi);
517
518 f2fs_put_page(dn->node_page, 1);
519 dn->node_page = NULL;
51dd6249 520 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
e05df3b1
JK		 521}
522
523static int truncate_dnode(struct dnode_of_data *dn)
524{
525 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
526 struct page *page;
527
528 if (dn->nid == 0)
529 return 1;
530
531 /* get direct node */
532 page = get_node_page(sbi, dn->nid);
533 if (IS_ERR(page) && PTR_ERR(page) == -ENOENT)
534 return 1;
535 else if (IS_ERR(page))
536 return PTR_ERR(page);
537
538 /* Make dnode_of_data for parameter */
539 dn->node_page = page;
540 dn->ofs_in_node = 0;
541 truncate_data_blocks(dn);
542 truncate_node(dn);
543 return 1;
544}
545
546static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
547 int ofs, int depth)
548{
549 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
550 struct dnode_of_data rdn = *dn;
551 struct page *page;
552 struct f2fs_node *rn;
553 nid_t child_nid;
554 unsigned int child_nofs;
555 int freed = 0;
556 int i, ret;
557
558 if (dn->nid == 0)
559 return NIDS_PER_BLOCK + 1;
560
51dd6249
NJ		 561 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
562
e05df3b1 563 page = get_node_page(sbi, dn->nid);
51dd6249
NJ		 564 if (IS_ERR(page)) {
565 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
e05df3b1 566 return PTR_ERR(page);
51dd6249 567 }
e05df3b1 568
45590710 569 rn = F2FS_NODE(page);
e05df3b1
JK		 570 if (depth < 3) {
571 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
572 child_nid = le32_to_cpu(rn->in.nid[i]);
573 if (child_nid == 0)
574 continue;
575 rdn.nid = child_nid;
576 ret = truncate_dnode(&rdn);
577 if (ret < 0)
578 goto out_err;
579 set_nid(page, i, 0, false);
580 }
581 } else {
582 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
583 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
584 child_nid = le32_to_cpu(rn->in.nid[i]);
585 if (child_nid == 0) {
586 child_nofs += NIDS_PER_BLOCK + 1;
587 continue;
588 }
589 rdn.nid = child_nid;
590 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
591 if (ret == (NIDS_PER_BLOCK + 1)) {
592 set_nid(page, i, 0, false);
593 child_nofs += ret;
594 } else if (ret < 0 && ret != -ENOENT) {
595 goto out_err;
596 }
597 }
598 freed = child_nofs;
599 }
600
601 if (!ofs) {
602 /* remove current indirect node */
603 dn->node_page = page;
604 truncate_node(dn);
605 freed++;
606 } else {
607 f2fs_put_page(page, 1);
608 }
51dd6249 609 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
e05df3b1
JK		 610 return freed;
611
612out_err:
613 f2fs_put_page(page, 1);
51dd6249 614 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
e05df3b1
JK		 615 return ret;
616}
617
618static int truncate_partial_nodes(struct dnode_of_data *dn,
619 struct f2fs_inode *ri, int *offset, int depth)
620{
621 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
622 struct page *pages[2];
623 nid_t nid[3];
624 nid_t child_nid;
625 int err = 0;
626 int i;
627 int idx = depth - 2;
628
629 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
630 if (!nid[0])
631 return 0;
632
633 /* get indirect nodes in the path */
634 for (i = 0; i < depth - 1; i++) {
635 /* refernece count'll be increased */
636 pages[i] = get_node_page(sbi, nid[i]);
637 if (IS_ERR(pages[i])) {
638 depth = i + 1;
639 err = PTR_ERR(pages[i]);
640 goto fail;
641 }
642 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
643 }
644
645 /* free direct nodes linked to a partial indirect node */
646 for (i = offset[depth - 1]; i < NIDS_PER_BLOCK; i++) {
647 child_nid = get_nid(pages[idx], i, false);
648 if (!child_nid)
649 continue;
650 dn->nid = child_nid;
651 err = truncate_dnode(dn);
652 if (err < 0)
653 goto fail;
654 set_nid(pages[idx], i, 0, false);
655 }
656
657 if (offset[depth - 1] == 0) {
658 dn->node_page = pages[idx];
659 dn->nid = nid[idx];
660 truncate_node(dn);
661 } else {
662 f2fs_put_page(pages[idx], 1);
663 }
664 offset[idx]++;
665 offset[depth - 1] = 0;
666fail:
667 for (i = depth - 3; i >= 0; i--)
668 f2fs_put_page(pages[i], 1);
51dd6249
NJ		 669
670 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
671
e05df3b1
JK		 672 return err;
673}
674
0a8165d7 675/*
e05df3b1
JK		 676 * All the block addresses of data and nodes should be nullified.
677 */
678int truncate_inode_blocks(struct inode *inode, pgoff_t from)
679{
680 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
afcb7ca0 681 struct address_space *node_mapping = sbi->node_inode->i_mapping;
e05df3b1
JK		 682 int err = 0, cont = 1;
683 int level, offset[4], noffset[4];
7dd690c8 684 unsigned int nofs = 0;
e05df3b1
JK		 685 struct f2fs_node *rn;
686 struct dnode_of_data dn;
687 struct page *page;
688
51dd6249
NJ		 689 trace_f2fs_truncate_inode_blocks_enter(inode, from);
690
de93653f 691 level = get_node_path(F2FS_I(inode), from, offset, noffset);
afcb7ca0 692restart:
e05df3b1 693 page = get_node_page(sbi, inode->i_ino);
51dd6249
NJ		 694 if (IS_ERR(page)) {
695 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
e05df3b1 696 return PTR_ERR(page);
51dd6249 697 }
e05df3b1
JK		 698
699 set_new_dnode(&dn, inode, page, NULL, 0);
700 unlock_page(page);
701
45590710 702 rn = F2FS_NODE(page);
e05df3b1
JK		 703 switch (level) {
704 case 0:
705 case 1:
706 nofs = noffset[1];
707 break;
708 case 2:
709 nofs = noffset[1];
710 if (!offset[level - 1])
711 goto skip_partial;
712 err = truncate_partial_nodes(&dn, &rn->i, offset, level);
713 if (err < 0 && err != -ENOENT)
714 goto fail;
715 nofs += 1 + NIDS_PER_BLOCK;
716 break;
717 case 3:
718 nofs = 5 + 2 * NIDS_PER_BLOCK;
719 if (!offset[level - 1])
720 goto skip_partial;
721 err = truncate_partial_nodes(&dn, &rn->i, offset, level);
722 if (err < 0 && err != -ENOENT)
723 goto fail;
724 break;
725 default:
726 BUG();
727 }
728
729skip_partial:
730 while (cont) {
731 dn.nid = le32_to_cpu(rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK]);
732 switch (offset[0]) {
733 case NODE_DIR1_BLOCK:
734 case NODE_DIR2_BLOCK:
735 err = truncate_dnode(&dn);
736 break;
737
738 case NODE_IND1_BLOCK:
739 case NODE_IND2_BLOCK:
740 err = truncate_nodes(&dn, nofs, offset[1], 2);
741 break;
742
743 case NODE_DIND_BLOCK:
744 err = truncate_nodes(&dn, nofs, offset[1], 3);
745 cont = 0;
746 break;
747
748 default:
749 BUG();
750 }
751 if (err < 0 && err != -ENOENT)
752 goto fail;
753 if (offset[1] == 0 &&
754 rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK]) {
755 lock_page(page);
afcb7ca0
JK		 756 if (page->mapping != node_mapping) {
757 f2fs_put_page(page, 1);
758 goto restart;
759 }
e05df3b1
JK		 760 wait_on_page_writeback(page);
761 rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
762 set_page_dirty(page);
763 unlock_page(page);
764 }
765 offset[1] = 0;
766 offset[0]++;
767 nofs += err;
768 }
769fail:
770 f2fs_put_page(page, 0);
51dd6249 771 trace_f2fs_truncate_inode_blocks_exit(inode, err);
e05df3b1
JK		 772 return err > 0 ? 0 : err;
773}
774
4f16fb0f
JK		 775int truncate_xattr_node(struct inode *inode, struct page *page)
776{
777 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
778 nid_t nid = F2FS_I(inode)->i_xattr_nid;
779 struct dnode_of_data dn;
780 struct page *npage;
781
782 if (!nid)
783 return 0;
784
785 npage = get_node_page(sbi, nid);
786 if (IS_ERR(npage))
787 return PTR_ERR(npage);
788
789 F2FS_I(inode)->i_xattr_nid = 0;
65985d93
JK		 790
791 /* need to do checkpoint during fsync */
792 F2FS_I(inode)->xattr_ver = cur_cp_version(F2FS_CKPT(sbi));
793
4f16fb0f
JK		 794 set_new_dnode(&dn, inode, page, npage, nid);
795
796 if (page)
797 dn.inode_page_locked = 1;
798 truncate_node(&dn);
799 return 0;
800}
801
39936837
JK		 802/*
803 * Caller should grab and release a mutex by calling mutex_lock_op() and
804 * mutex_unlock_op().
805 */
e05df3b1
JK		 806int remove_inode_page(struct inode *inode)
807{
808 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
809 struct page *page;
810 nid_t ino = inode->i_ino;
811 struct dnode_of_data dn;
4f16fb0f 812 int err;
e05df3b1 813
e05df3b1 814 page = get_node_page(sbi, ino);
39936837 815 if (IS_ERR(page))
e05df3b1 816 return PTR_ERR(page);
e05df3b1 817
4f16fb0f
JK		 818 err = truncate_xattr_node(inode, page);
819 if (err) {
820 f2fs_put_page(page, 1);
821 return err;
e05df3b1 822 }
e05df3b1 823
71e9fec5 824 /* 0 is possible, after f2fs_new_inode() is failed */
5d56b671 825 f2fs_bug_on(inode->i_blocks != 0 && inode->i_blocks != 1);
71e9fec5
JK		 826 set_new_dnode(&dn, inode, page, page, ino);
827 truncate_node(&dn);
e05df3b1
JK		 828 return 0;
829}
830
44a83ff6 831struct page *new_inode_page(struct inode *inode, const struct qstr *name)
e05df3b1 832{
e05df3b1
JK		 833 struct dnode_of_data dn;
834
835 /* allocate inode page for new inode */
836 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
44a83ff6
JK		 837
838 /* caller should f2fs_put_page(page, 1); */
8ae8f162 839 return new_node_page(&dn, 0, NULL);
e05df3b1
JK		 840}
841
8ae8f162
JK		 842struct page *new_node_page(struct dnode_of_data *dn,
843 unsigned int ofs, struct page *ipage)
e05df3b1
JK		 844{
845 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
846 struct address_space *mapping = sbi->node_inode->i_mapping;
847 struct node_info old_ni, new_ni;
848 struct page *page;
849 int err;
850
851 if (is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))
852 return ERR_PTR(-EPERM);
853
854 page = grab_cache_page(mapping, dn->nid);
855 if (!page)
856 return ERR_PTR(-ENOMEM);
857
9c02740c
JK		 858 if (!inc_valid_node_count(sbi, dn->inode, 1)) {
859 err = -ENOSPC;
860 goto fail;
861 }
e05df3b1 862
9c02740c 863 get_node_info(sbi, dn->nid, &old_ni);
e05df3b1
JK		 864
865 /* Reinitialize old_ni with new node page */
5d56b671 866 f2fs_bug_on(old_ni.blk_addr != NULL_ADDR);
e05df3b1
JK		 867 new_ni = old_ni;
868 new_ni.ino = dn->inode->i_ino;
e05df3b1 869 set_node_addr(sbi, &new_ni, NEW_ADDR);
9c02740c
JK		 870
871 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
398b1ac5 872 set_cold_node(dn->inode, page);
9c02740c
JK		 873 SetPageUptodate(page);
874 set_page_dirty(page);
e05df3b1 875
479bd73a
JK		 876 if (ofs == XATTR_NODE_OFFSET)
877 F2FS_I(dn->inode)->i_xattr_nid = dn->nid;
878
e05df3b1 879 dn->node_page = page;
8ae8f162
JK		 880 if (ipage)
881 update_inode(dn->inode, ipage);
882 else
883 sync_inode_page(dn);
e05df3b1
JK		 884 if (ofs == 0)
885 inc_valid_inode_count(sbi);
886
887 return page;
888
889fail:
71e9fec5 890 clear_node_page_dirty(page);
e05df3b1
JK		 891 f2fs_put_page(page, 1);
892 return ERR_PTR(err);
893}
894
56ae674c
JK		 895/*
896 * Caller should do after getting the following values.
897 * 0: f2fs_put_page(page, 0)
898 * LOCKED_PAGE: f2fs_put_page(page, 1)
899 * error: nothing
900 */
e05df3b1
JK		 901static int read_node_page(struct page *page, int type)
902{
903 struct f2fs_sb_info *sbi = F2FS_SB(page->mapping->host->i_sb);
904 struct node_info ni;
905
906 get_node_info(sbi, page->index, &ni);
907
393ff91f
JK		 908 if (ni.blk_addr == NULL_ADDR) {
909 f2fs_put_page(page, 1);
e05df3b1 910 return -ENOENT;
393ff91f
JK		 911 }
912
56ae674c
JK		 913 if (PageUptodate(page))
914 return LOCKED_PAGE;
393ff91f 915
e05df3b1
JK		 916 return f2fs_readpage(sbi, page, ni.blk_addr, type);
917}
918
0a8165d7 919/*
e05df3b1
JK		 920 * Readahead a node page
921 */
922void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
923{
924 struct address_space *mapping = sbi->node_inode->i_mapping;
925 struct page *apage;
56ae674c 926 int err;
e05df3b1
JK		 927
928 apage = find_get_page(mapping, nid);
393ff91f
JK		 929 if (apage && PageUptodate(apage)) {
930 f2fs_put_page(apage, 0);
931 return;
932 }
e05df3b1
JK		 933 f2fs_put_page(apage, 0);
934
935 apage = grab_cache_page(mapping, nid);
936 if (!apage)
937 return;
938
56ae674c
JK		 939 err = read_node_page(apage, READA);
940 if (err == 0)
393ff91f 941 f2fs_put_page(apage, 0);
56ae674c
JK		 942 else if (err == LOCKED_PAGE)
943 f2fs_put_page(apage, 1);
e05df3b1
JK		 944}
945
946struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
947{
e05df3b1 948 struct address_space *mapping = sbi->node_inode->i_mapping;
56ae674c
JK		 949 struct page *page;
950 int err;
afcb7ca0 951repeat:
e05df3b1
JK		 952 page = grab_cache_page(mapping, nid);
953 if (!page)
954 return ERR_PTR(-ENOMEM);
955
956 err = read_node_page(page, READ_SYNC);
56ae674c 957 if (err < 0)
e05df3b1 958 return ERR_PTR(err);
56ae674c
JK		 959 else if (err == LOCKED_PAGE)
960 goto got_it;
e05df3b1 961
393ff91f
JK		 962 lock_page(page);
963 if (!PageUptodate(page)) {
964 f2fs_put_page(page, 1);
965 return ERR_PTR(-EIO);
966 }
afcb7ca0
JK		 967 if (page->mapping != mapping) {
968 f2fs_put_page(page, 1);
969 goto repeat;
970 }
56ae674c 971got_it:
5d56b671 972 f2fs_bug_on(nid != nid_of_node(page));
e05df3b1
JK		 973 mark_page_accessed(page);
974 return page;
975}
976
0a8165d7 977/*
e05df3b1
JK		 978 * Return a locked page for the desired node page.
979 * And, readahead MAX_RA_NODE number of node pages.
980 */
981struct page *get_node_page_ra(struct page *parent, int start)
982{
983 struct f2fs_sb_info *sbi = F2FS_SB(parent->mapping->host->i_sb);
984 struct address_space *mapping = sbi->node_inode->i_mapping;
c718379b 985 struct blk_plug plug;
e05df3b1 986 struct page *page;
56ae674c
JK		 987 int err, i, end;
988 nid_t nid;
e05df3b1
JK		 989
990 /* First, try getting the desired direct node. */
991 nid = get_nid(parent, start, false);
992 if (!nid)
993 return ERR_PTR(-ENOENT);
afcb7ca0 994repeat:
e05df3b1
JK		 995 page = grab_cache_page(mapping, nid);
996 if (!page)
997 return ERR_PTR(-ENOMEM);
998
66d36a29 999 err = read_node_page(page, READ_SYNC);
56ae674c 1000 if (err < 0)
e05df3b1 1001 return ERR_PTR(err);
56ae674c
JK		 1002 else if (err == LOCKED_PAGE)
1003 goto page_hit;
e05df3b1 1004
c718379b
JK		 1005 blk_start_plug(&plug);
1006
e05df3b1
JK		 1007 /* Then, try readahead for siblings of the desired node */
1008 end = start + MAX_RA_NODE;
1009 end = min(end, NIDS_PER_BLOCK);
1010 for (i = start + 1; i < end; i++) {
1011 nid = get_nid(parent, i, false);
1012 if (!nid)
1013 continue;
1014 ra_node_page(sbi, nid);
1015 }
1016
c718379b
JK		 1017 blk_finish_plug(&plug);
1018
e05df3b1 1019 lock_page(page);
afcb7ca0
JK		 1020 if (page->mapping != mapping) {
1021 f2fs_put_page(page, 1);
1022 goto repeat;
1023 }
e0f56cb4 1024page_hit:
56ae674c 1025 if (!PageUptodate(page)) {
e05df3b1
JK		 1026 f2fs_put_page(page, 1);
1027 return ERR_PTR(-EIO);
1028 }
393ff91f 1029 mark_page_accessed(page);
e05df3b1
JK		 1030 return page;
1031}
1032
1033void sync_inode_page(struct dnode_of_data *dn)
1034{
1035 if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) {
1036 update_inode(dn->inode, dn->node_page);
1037 } else if (dn->inode_page) {
1038 if (!dn->inode_page_locked)
1039 lock_page(dn->inode_page);
1040 update_inode(dn->inode, dn->inode_page);
1041 if (!dn->inode_page_locked)
1042 unlock_page(dn->inode_page);
1043 } else {
39936837 1044 update_inode_page(dn->inode);
e05df3b1
JK		 1045 }
1046}
1047
1048int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino,
1049 struct writeback_control *wbc)
1050{
1051 struct address_space *mapping = sbi->node_inode->i_mapping;
1052 pgoff_t index, end;
1053 struct pagevec pvec;
1054 int step = ino ? 2 : 0;
1055 int nwritten = 0, wrote = 0;
1056
1057 pagevec_init(&pvec, 0);
1058
1059next_step:
1060 index = 0;
1061 end = LONG_MAX;
1062
1063 while (index <= end) {
1064 int i, nr_pages;
1065 nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
1066 PAGECACHE_TAG_DIRTY,
1067 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
1068 if (nr_pages == 0)
1069 break;
1070
1071 for (i = 0; i < nr_pages; i++) {
1072 struct page *page = pvec.pages[i];
1073
1074 /*
1075 * flushing sequence with step:
1076 * 0. indirect nodes
1077 * 1. dentry dnodes
1078 * 2. file dnodes
1079 */
1080 if (step == 0 && IS_DNODE(page))
1081 continue;
1082 if (step == 1 && (!IS_DNODE(page) ||
1083 is_cold_node(page)))
1084 continue;
1085 if (step == 2 && (!IS_DNODE(page) ||
1086 !is_cold_node(page)))
1087 continue;
1088
1089 /*
1090 * If an fsync mode,
1091 * we should not skip writing node pages.
1092 */
1093 if (ino && ino_of_node(page) == ino)
1094 lock_page(page);
1095 else if (!trylock_page(page))
1096 continue;
1097
1098 if (unlikely(page->mapping != mapping)) {
1099continue_unlock:
1100 unlock_page(page);
1101 continue;
1102 }
1103 if (ino && ino_of_node(page) != ino)
1104 goto continue_unlock;
1105
1106 if (!PageDirty(page)) {
1107 /* someone wrote it for us */
1108 goto continue_unlock;
1109 }
1110
1111 if (!clear_page_dirty_for_io(page))
1112 goto continue_unlock;
1113
1114 /* called by fsync() */
1115 if (ino && IS_DNODE(page)) {
1116 int mark = !is_checkpointed_node(sbi, ino);
1117 set_fsync_mark(page, 1);
1118 if (IS_INODE(page))
1119 set_dentry_mark(page, mark);
1120 nwritten++;
1121 } else {
1122 set_fsync_mark(page, 0);
1123 set_dentry_mark(page, 0);
1124 }
1125 mapping->a_ops->writepage(page, wbc);
1126 wrote++;
1127
1128 if (--wbc->nr_to_write == 0)
1129 break;
1130 }
1131 pagevec_release(&pvec);
1132 cond_resched();
1133
1134 if (wbc->nr_to_write == 0) {
1135 step = 2;
1136 break;
1137 }
1138 }
1139
1140 if (step < 2) {
1141 step++;
1142 goto next_step;
1143 }
1144
1145 if (wrote)
1146 f2fs_submit_bio(sbi, NODE, wbc->sync_mode == WB_SYNC_ALL);
1147
1148 return nwritten;
1149}
1150
cfe58f9d
JK		 1151int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino)
1152{
1153 struct address_space *mapping = sbi->node_inode->i_mapping;
1154 pgoff_t index = 0, end = LONG_MAX;
1155 struct pagevec pvec;
1156 int nr_pages;
1157 int ret2 = 0, ret = 0;
1158
1159 pagevec_init(&pvec, 0);
1160 while ((index <= end) &&
1161 (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
1162 PAGECACHE_TAG_WRITEBACK,
1163 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1)) != 0) {
1164 unsigned i;
1165
1166 for (i = 0; i < nr_pages; i++) {
1167 struct page *page = pvec.pages[i];
1168
1169 /* until radix tree lookup accepts end_index */
1170 if (page->index > end)
1171 continue;
1172
1173 if (ino && ino_of_node(page) == ino)
1174 wait_on_page_writeback(page);
1175 if (TestClearPageError(page))
1176 ret = -EIO;
1177 }
1178 pagevec_release(&pvec);
1179 cond_resched();
1180 }
1181
1182 if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
1183 ret2 = -ENOSPC;
1184 if (test_and_clear_bit(AS_EIO, &mapping->flags))
1185 ret2 = -EIO;
1186 if (!ret)
1187 ret = ret2;
1188 return ret;
1189}
1190
e05df3b1
JK		 1191static int f2fs_write_node_page(struct page *page,
1192 struct writeback_control *wbc)
1193{
1194 struct f2fs_sb_info *sbi = F2FS_SB(page->mapping->host->i_sb);
1195 nid_t nid;
e05df3b1
JK		 1196 block_t new_addr;
1197 struct node_info ni;
1198
87a9bd26
JK		 1199 if (sbi->por_doing)
1200 goto redirty_out;
1201
e05df3b1
JK		 1202 wait_on_page_writeback(page);
1203
e05df3b1
JK		 1204 /* get old block addr of this node page */
1205 nid = nid_of_node(page);
5d56b671 1206 f2fs_bug_on(page->index != nid);
e05df3b1
JK		 1207
1208 get_node_info(sbi, nid, &ni);
1209
1210 /* This page is already truncated */
39936837
JK		 1211 if (ni.blk_addr == NULL_ADDR) {
1212 dec_page_count(sbi, F2FS_DIRTY_NODES);
1213 unlock_page(page);
1214 return 0;
1215 }
e05df3b1 1216
87a9bd26
JK		 1217 if (wbc->for_reclaim)
1218 goto redirty_out;
08d8058b 1219
39936837 1220 mutex_lock(&sbi->node_write);
e05df3b1 1221 set_page_writeback(page);
e05df3b1
JK		 1222 write_node_page(sbi, page, nid, ni.blk_addr, &new_addr);
1223 set_node_addr(sbi, &ni, new_addr);
1224 dec_page_count(sbi, F2FS_DIRTY_NODES);
39936837 1225 mutex_unlock(&sbi->node_write);
e05df3b1
JK		 1226 unlock_page(page);
1227 return 0;
87a9bd26
JK		 1228
1229redirty_out:
1230 dec_page_count(sbi, F2FS_DIRTY_NODES);
1231 wbc->pages_skipped++;
1232 set_page_dirty(page);
1233 return AOP_WRITEPAGE_ACTIVATE;
e05df3b1
JK		 1234}
1235
a7fdffbd
JK		 1236/*
1237 * It is very important to gather dirty pages and write at once, so that we can
1238 * submit a big bio without interfering other data writes.
423e95cc 1239 * Be default, 512 pages (2MB) * 3 node types, is more reasonable.
a7fdffbd 1240 */
423e95cc 1241#define COLLECT_DIRTY_NODES 1536
e05df3b1
JK		 1242static int f2fs_write_node_pages(struct address_space *mapping,
1243 struct writeback_control *wbc)
1244{
1245 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
e05df3b1
JK		 1246 long nr_to_write = wbc->nr_to_write;
1247
4660f9c0
JK		 1248 /* balancing f2fs's metadata in background */
1249 f2fs_balance_fs_bg(sbi);
e05df3b1 1250
a7fdffbd
JK		 1251 /* collect a number of dirty node pages and write together */
1252 if (get_pages(sbi, F2FS_DIRTY_NODES) < COLLECT_DIRTY_NODES)
1253 return 0;
1254
e05df3b1 1255 /* if mounting is failed, skip writing node pages */
423e95cc 1256 wbc->nr_to_write = 3 * max_hw_blocks(sbi);
e05df3b1 1257 sync_node_pages(sbi, 0, wbc);
423e95cc
JK		 1258 wbc->nr_to_write = nr_to_write - (3 * max_hw_blocks(sbi) -
1259 wbc->nr_to_write);
e05df3b1
JK		 1260 return 0;
1261}
1262
1263static int f2fs_set_node_page_dirty(struct page *page)
1264{
1265 struct address_space *mapping = page->mapping;
1266 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb);
1267
26c6b887
JK		 1268 trace_f2fs_set_page_dirty(page, NODE);
1269
e05df3b1
JK		 1270 SetPageUptodate(page);
1271 if (!PageDirty(page)) {
1272 __set_page_dirty_nobuffers(page);
1273 inc_page_count(sbi, F2FS_DIRTY_NODES);
1274 SetPagePrivate(page);
1275 return 1;
1276 }
1277 return 0;
1278}
1279
d47992f8
LC		 1280static void f2fs_invalidate_node_page(struct page *page, unsigned int offset,
1281 unsigned int length)
e05df3b1
JK		 1282{
1283 struct inode *inode = page->mapping->host;
1284 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
1285 if (PageDirty(page))
1286 dec_page_count(sbi, F2FS_DIRTY_NODES);
1287 ClearPagePrivate(page);
1288}
1289
1290static int f2fs_release_node_page(struct page *page, gfp_t wait)
1291{
1292 ClearPagePrivate(page);
c3850aa1 1293 return 1;
e05df3b1
JK		 1294}
1295
0a8165d7 1296/*
e05df3b1
JK		 1297 * Structure of the f2fs node operations
1298 */
1299const struct address_space_operations f2fs_node_aops = {
1300 .writepage = f2fs_write_node_page,
1301 .writepages = f2fs_write_node_pages,
1302 .set_page_dirty = f2fs_set_node_page_dirty,
1303 .invalidatepage = f2fs_invalidate_node_page,
1304 .releasepage = f2fs_release_node_page,
1305};
1306
1307static struct free_nid *__lookup_free_nid_list(nid_t n, struct list_head *head)
1308{
1309 struct list_head *this;
3aa770a9 1310 struct free_nid *i;
e05df3b1
JK		 1311 list_for_each(this, head) {
1312 i = list_entry(this, struct free_nid, list);
1313 if (i->nid == n)
3aa770a9 1314 return i;
e05df3b1 1315 }
3aa770a9 1316 return NULL;
e05df3b1
JK		 1317}
1318
1319static void __del_from_free_nid_list(struct free_nid *i)
1320{
1321 list_del(&i->list);
1322 kmem_cache_free(free_nid_slab, i);
1323}
1324
59bbd474 1325static int add_free_nid(struct f2fs_nm_info *nm_i, nid_t nid, bool build)
e05df3b1
JK		 1326{
1327 struct free_nid *i;
59bbd474
JK		 1328 struct nat_entry *ne;
1329 bool allocated = false;
e05df3b1
JK		 1330
1331 if (nm_i->fcnt > 2 * MAX_FREE_NIDS)
23d38844 1332 return -1;
9198aceb
JK		 1333
1334 /* 0 nid should not be used */
1335 if (nid == 0)
1336 return 0;
59bbd474 1337
7bd59381
GZ		 1338 if (build) {
1339 /* do not add allocated nids */
1340 read_lock(&nm_i->nat_tree_lock);
1341 ne = __lookup_nat_cache(nm_i, nid);
1342 if (ne && nat_get_blkaddr(ne) != NULL_ADDR)
1343 allocated = true;
1344 read_unlock(&nm_i->nat_tree_lock);
1345 if (allocated)
1346 return 0;
e05df3b1 1347 }
7bd59381
GZ		 1348
1349 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
e05df3b1
JK		 1350 i->nid = nid;
1351 i->state = NID_NEW;
1352
1353 spin_lock(&nm_i->free_nid_list_lock);
1354 if (__lookup_free_nid_list(nid, &nm_i->free_nid_list)) {
1355 spin_unlock(&nm_i->free_nid_list_lock);
1356 kmem_cache_free(free_nid_slab, i);
1357 return 0;
1358 }
1359 list_add_tail(&i->list, &nm_i->free_nid_list);
1360 nm_i->fcnt++;
1361 spin_unlock(&nm_i->free_nid_list_lock);
1362 return 1;
1363}
1364
1365static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid)
1366{
1367 struct free_nid *i;
1368 spin_lock(&nm_i->free_nid_list_lock);
1369 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list);
1370 if (i && i->state == NID_NEW) {
1371 __del_from_free_nid_list(i);
1372 nm_i->fcnt--;
1373 }
1374 spin_unlock(&nm_i->free_nid_list_lock);
1375}
1376
8760952d 1377static void scan_nat_page(struct f2fs_nm_info *nm_i,
e05df3b1
JK		 1378 struct page *nat_page, nid_t start_nid)
1379{
1380 struct f2fs_nat_block *nat_blk = page_address(nat_page);
1381 block_t blk_addr;
e05df3b1
JK		 1382 int i;
1383
e05df3b1
JK		 1384 i = start_nid % NAT_ENTRY_PER_BLOCK;
1385
1386 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
23d38844 1387
04431c44
JK		 1388 if (start_nid >= nm_i->max_nid)
1389 break;
23d38844
HL		 1390
1391 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
5d56b671 1392 f2fs_bug_on(blk_addr == NEW_ADDR);
23d38844 1393 if (blk_addr == NULL_ADDR) {
59bbd474 1394 if (add_free_nid(nm_i, start_nid, true) < 0)
23d38844
HL		 1395 break;
1396 }
e05df3b1 1397 }
e05df3b1
JK		 1398}
1399
1400static void build_free_nids(struct f2fs_sb_info *sbi)
1401{
e05df3b1
JK		 1402 struct f2fs_nm_info *nm_i = NM_I(sbi);
1403 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1404 struct f2fs_summary_block *sum = curseg->sum_blk;
8760952d 1405 int i = 0;
55008d84 1406 nid_t nid = nm_i->next_scan_nid;
e05df3b1 1407
55008d84
JK		 1408 /* Enough entries */
1409 if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK)
1410 return;
e05df3b1 1411
55008d84 1412 /* readahead nat pages to be scanned */
e05df3b1
JK		 1413 ra_nat_pages(sbi, nid);
1414
1415 while (1) {
1416 struct page *page = get_current_nat_page(sbi, nid);
1417
8760952d 1418 scan_nat_page(nm_i, page, nid);
e05df3b1
JK		 1419 f2fs_put_page(page, 1);
1420
1421 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
55008d84 1422 if (nid >= nm_i->max_nid)
e05df3b1 1423 nid = 0;
55008d84
JK		 1424
1425 if (i++ == FREE_NID_PAGES)
e05df3b1
JK		 1426 break;
1427 }
1428
55008d84
JK		 1429 /* go to the next free nat pages to find free nids abundantly */
1430 nm_i->next_scan_nid = nid;
e05df3b1
JK		 1431
1432 /* find free nids from current sum_pages */
1433 mutex_lock(&curseg->curseg_mutex);
1434 for (i = 0; i < nats_in_cursum(sum); i++) {
1435 block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr);
1436 nid = le32_to_cpu(nid_in_journal(sum, i));
1437 if (addr == NULL_ADDR)
59bbd474 1438 add_free_nid(nm_i, nid, true);
e05df3b1
JK		 1439 else
1440 remove_free_nid(nm_i, nid);
1441 }
1442 mutex_unlock(&curseg->curseg_mutex);
e05df3b1
JK		 1443}
1444
1445/*
1446 * If this function returns success, caller can obtain a new nid
1447 * from second parameter of this function.
1448 * The returned nid could be used ino as well as nid when inode is created.
1449 */
1450bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
1451{
1452 struct f2fs_nm_info *nm_i = NM_I(sbi);
1453 struct free_nid *i = NULL;
1454 struct list_head *this;
1455retry:
55008d84
JK		 1456 if (sbi->total_valid_node_count + 1 >= nm_i->max_nid)
1457 return false;
e05df3b1 1458
e05df3b1 1459 spin_lock(&nm_i->free_nid_list_lock);
e05df3b1 1460
55008d84
JK		 1461 /* We should not use stale free nids created by build_free_nids */
1462 if (nm_i->fcnt && !sbi->on_build_free_nids) {
5d56b671 1463 f2fs_bug_on(list_empty(&nm_i->free_nid_list));
55008d84
JK		 1464 list_for_each(this, &nm_i->free_nid_list) {
1465 i = list_entry(this, struct free_nid, list);
1466 if (i->state == NID_NEW)
1467 break;
1468 }
e05df3b1 1469
5d56b671 1470 f2fs_bug_on(i->state != NID_NEW);
55008d84
JK		 1471 *nid = i->nid;
1472 i->state = NID_ALLOC;
1473 nm_i->fcnt--;
1474 spin_unlock(&nm_i->free_nid_list_lock);
1475 return true;
1476 }
e05df3b1 1477 spin_unlock(&nm_i->free_nid_list_lock);
55008d84
JK		 1478
1479 /* Let's scan nat pages and its caches to get free nids */
1480 mutex_lock(&nm_i->build_lock);
aabe5136 1481 sbi->on_build_free_nids = true;
55008d84 1482 build_free_nids(sbi);
aabe5136 1483 sbi->on_build_free_nids = false;
55008d84
JK		 1484 mutex_unlock(&nm_i->build_lock);
1485 goto retry;
e05df3b1
JK		 1486}
1487
0a8165d7 1488/*
e05df3b1
JK		 1489 * alloc_nid() should be called prior to this function.
1490 */
1491void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
1492{
1493 struct f2fs_nm_info *nm_i = NM_I(sbi);
1494 struct free_nid *i;
1495
1496 spin_lock(&nm_i->free_nid_list_lock);
1497 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list);
5d56b671 1498 f2fs_bug_on(!i || i->state != NID_ALLOC);
49952fa1 1499 __del_from_free_nid_list(i);
e05df3b1
JK		 1500 spin_unlock(&nm_i->free_nid_list_lock);
1501}
1502
0a8165d7 1503/*
e05df3b1
JK		 1504 * alloc_nid() should be called prior to this function.
1505 */
1506void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
1507{
49952fa1
JK		 1508 struct f2fs_nm_info *nm_i = NM_I(sbi);
1509 struct free_nid *i;
1510
65985d93
JK		 1511 if (!nid)
1512 return;
1513
49952fa1
JK		 1514 spin_lock(&nm_i->free_nid_list_lock);
1515 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list);
5d56b671 1516 f2fs_bug_on(!i || i->state != NID_ALLOC);
95630cba
HL		 1517 if (nm_i->fcnt > 2 * MAX_FREE_NIDS) {
1518 __del_from_free_nid_list(i);
1519 } else {
1520 i->state = NID_NEW;
1521 nm_i->fcnt++;
1522 }
49952fa1 1523 spin_unlock(&nm_i->free_nid_list_lock);
e05df3b1
JK		 1524}
1525
1526void recover_node_page(struct f2fs_sb_info *sbi, struct page *page,
1527 struct f2fs_summary *sum, struct node_info *ni,
1528 block_t new_blkaddr)
1529{
1530 rewrite_node_page(sbi, page, sum, ni->blk_addr, new_blkaddr);
1531 set_node_addr(sbi, ni, new_blkaddr);
1532 clear_node_page_dirty(page);
1533}
1534
1535int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
1536{
1537 struct address_space *mapping = sbi->node_inode->i_mapping;
1538 struct f2fs_node *src, *dst;
1539 nid_t ino = ino_of_node(page);
1540 struct node_info old_ni, new_ni;
1541 struct page *ipage;
1542
1543 ipage = grab_cache_page(mapping, ino);
1544 if (!ipage)
1545 return -ENOMEM;
1546
1547 /* Should not use this inode from free nid list */
1548 remove_free_nid(NM_I(sbi), ino);
1549
1550 get_node_info(sbi, ino, &old_ni);
1551 SetPageUptodate(ipage);
1552 fill_node_footer(ipage, ino, ino, 0, true);
1553
45590710
GZ		 1554 src = F2FS_NODE(page);
1555 dst = F2FS_NODE(ipage);
e05df3b1
JK		 1556
1557 memcpy(dst, src, (unsigned long)&src->i.i_ext - (unsigned long)&src->i);
1558 dst->i.i_size = 0;
25ca923b
JK		 1559 dst->i.i_blocks = cpu_to_le64(1);
1560 dst->i.i_links = cpu_to_le32(1);
e05df3b1
JK		 1561 dst->i.i_xattr_nid = 0;
1562
1563 new_ni = old_ni;
1564 new_ni.ino = ino;
1565
65e5cd0a
JK		 1566 if (!inc_valid_node_count(sbi, NULL, 1))
1567 WARN_ON(1);
e05df3b1
JK		 1568 set_node_addr(sbi, &new_ni, NEW_ADDR);
1569 inc_valid_inode_count(sbi);
e05df3b1
JK		 1570 f2fs_put_page(ipage, 1);
1571 return 0;
1572}
1573
1574int restore_node_summary(struct f2fs_sb_info *sbi,
1575 unsigned int segno, struct f2fs_summary_block *sum)
1576{
1577 struct f2fs_node *rn;
1578 struct f2fs_summary *sum_entry;
1579 struct page *page;
1580 block_t addr;
1581 int i, last_offset;
1582
1583 /* alloc temporal page for read node */
1584 page = alloc_page(GFP_NOFS | __GFP_ZERO);
e27dae4d
DC		 1585 if (!page)
1586 return -ENOMEM;
e05df3b1
JK		 1587 lock_page(page);
1588
1589 /* scan the node segment */
1590 last_offset = sbi->blocks_per_seg;
1591 addr = START_BLOCK(sbi, segno);
1592 sum_entry = &sum->entries[0];
1593
1594 for (i = 0; i < last_offset; i++, sum_entry++) {
393ff91f
JK		 1595 /*
1596 * In order to read next node page,
1597 * we must clear PageUptodate flag.
1598 */
1599 ClearPageUptodate(page);
1600
e05df3b1
JK		 1601 if (f2fs_readpage(sbi, page, addr, READ_SYNC))
1602 goto out;
1603
393ff91f 1604 lock_page(page);
45590710 1605 rn = F2FS_NODE(page);
e05df3b1
JK		 1606 sum_entry->nid = rn->footer.nid;
1607 sum_entry->version = 0;
1608 sum_entry->ofs_in_node = 0;
1609 addr++;
e05df3b1 1610 }
e05df3b1 1611 unlock_page(page);
393ff91f 1612out:
e05df3b1
JK		 1613 __free_pages(page, 0);
1614 return 0;
1615}
1616
1617static bool flush_nats_in_journal(struct f2fs_sb_info *sbi)
1618{
1619 struct f2fs_nm_info *nm_i = NM_I(sbi);
1620 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1621 struct f2fs_summary_block *sum = curseg->sum_blk;
1622 int i;
1623
1624 mutex_lock(&curseg->curseg_mutex);
1625
1626 if (nats_in_cursum(sum) < NAT_JOURNAL_ENTRIES) {
1627 mutex_unlock(&curseg->curseg_mutex);
1628 return false;
1629 }
1630
1631 for (i = 0; i < nats_in_cursum(sum); i++) {
1632 struct nat_entry *ne;
1633 struct f2fs_nat_entry raw_ne;
1634 nid_t nid = le32_to_cpu(nid_in_journal(sum, i));
1635
1636 raw_ne = nat_in_journal(sum, i);
1637retry:
1638 write_lock(&nm_i->nat_tree_lock);
1639 ne = __lookup_nat_cache(nm_i, nid);
1640 if (ne) {
1641 __set_nat_cache_dirty(nm_i, ne);
1642 write_unlock(&nm_i->nat_tree_lock);
1643 continue;
1644 }
1645 ne = grab_nat_entry(nm_i, nid);
1646 if (!ne) {
1647 write_unlock(&nm_i->nat_tree_lock);
1648 goto retry;
1649 }
1650 nat_set_blkaddr(ne, le32_to_cpu(raw_ne.block_addr));
1651 nat_set_ino(ne, le32_to_cpu(raw_ne.ino));
1652 nat_set_version(ne, raw_ne.version);
1653 __set_nat_cache_dirty(nm_i, ne);
1654 write_unlock(&nm_i->nat_tree_lock);
1655 }
1656 update_nats_in_cursum(sum, -i);
1657 mutex_unlock(&curseg->curseg_mutex);
1658 return true;
1659}
1660
0a8165d7 1661/*
e05df3b1
JK		 1662 * This function is called during the checkpointing process.
1663 */
1664void flush_nat_entries(struct f2fs_sb_info *sbi)
1665{
1666 struct f2fs_nm_info *nm_i = NM_I(sbi);
1667 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1668 struct f2fs_summary_block *sum = curseg->sum_blk;
1669 struct list_head *cur, *n;
1670 struct page *page = NULL;
1671 struct f2fs_nat_block *nat_blk = NULL;
1672 nid_t start_nid = 0, end_nid = 0;
1673 bool flushed;
1674
1675 flushed = flush_nats_in_journal(sbi);
1676
1677 if (!flushed)
1678 mutex_lock(&curseg->curseg_mutex);
1679
1680 /* 1) flush dirty nat caches */
1681 list_for_each_safe(cur, n, &nm_i->dirty_nat_entries) {
1682 struct nat_entry *ne;
1683 nid_t nid;
1684 struct f2fs_nat_entry raw_ne;
1685 int offset = -1;
2b50638d 1686 block_t new_blkaddr;
e05df3b1
JK		 1687
1688 ne = list_entry(cur, struct nat_entry, list);
1689 nid = nat_get_nid(ne);
1690
1691 if (nat_get_blkaddr(ne) == NEW_ADDR)
1692 continue;
1693 if (flushed)
1694 goto to_nat_page;
1695
1696 /* if there is room for nat enries in curseg->sumpage */
1697 offset = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 1);
1698 if (offset >= 0) {
1699 raw_ne = nat_in_journal(sum, offset);
e05df3b1
JK		 1700 goto flush_now;
1701 }
1702to_nat_page:
1703 if (!page || (start_nid > nid || nid > end_nid)) {
1704 if (page) {
1705 f2fs_put_page(page, 1);
1706 page = NULL;
1707 }
1708 start_nid = START_NID(nid);
1709 end_nid = start_nid + NAT_ENTRY_PER_BLOCK - 1;
1710
1711 /*
1712 * get nat block with dirty flag, increased reference
1713 * count, mapped and lock
1714 */
1715 page = get_next_nat_page(sbi, start_nid);
1716 nat_blk = page_address(page);
1717 }
1718
5d56b671 1719 f2fs_bug_on(!nat_blk);
e05df3b1 1720 raw_ne = nat_blk->entries[nid - start_nid];
e05df3b1
JK		 1721flush_now:
1722 new_blkaddr = nat_get_blkaddr(ne);
1723
1724 raw_ne.ino = cpu_to_le32(nat_get_ino(ne));
1725 raw_ne.block_addr = cpu_to_le32(new_blkaddr);
1726 raw_ne.version = nat_get_version(ne);
1727
1728 if (offset < 0) {
1729 nat_blk->entries[nid - start_nid] = raw_ne;
1730 } else {
1731 nat_in_journal(sum, offset) = raw_ne;
1732 nid_in_journal(sum, offset) = cpu_to_le32(nid);
1733 }
1734
fa372417 1735 if (nat_get_blkaddr(ne) == NULL_ADDR &&
59bbd474 1736 add_free_nid(NM_I(sbi), nid, false) <= 0) {
e05df3b1
JK		 1737 write_lock(&nm_i->nat_tree_lock);
1738 __del_from_nat_cache(nm_i, ne);
1739 write_unlock(&nm_i->nat_tree_lock);
e05df3b1
JK		 1740 } else {
1741 write_lock(&nm_i->nat_tree_lock);
1742 __clear_nat_cache_dirty(nm_i, ne);
1743 ne->checkpointed = true;
1744 write_unlock(&nm_i->nat_tree_lock);
1745 }
1746 }
1747 if (!flushed)
1748 mutex_unlock(&curseg->curseg_mutex);
1749 f2fs_put_page(page, 1);
1750
1751 /* 2) shrink nat caches if necessary */
1752 try_to_free_nats(sbi, nm_i->nat_cnt - NM_WOUT_THRESHOLD);
1753}
1754
1755static int init_node_manager(struct f2fs_sb_info *sbi)
1756{
1757 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
1758 struct f2fs_nm_info *nm_i = NM_I(sbi);
1759 unsigned char *version_bitmap;
1760 unsigned int nat_segs, nat_blocks;
1761
1762 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
1763
1764 /* segment_count_nat includes pair segment so divide to 2. */
1765 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
1766 nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
1767 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks;
1768 nm_i->fcnt = 0;
1769 nm_i->nat_cnt = 0;
1770
1771 INIT_LIST_HEAD(&nm_i->free_nid_list);
1772 INIT_RADIX_TREE(&nm_i->nat_root, GFP_ATOMIC);
1773 INIT_LIST_HEAD(&nm_i->nat_entries);
1774 INIT_LIST_HEAD(&nm_i->dirty_nat_entries);
1775
1776 mutex_init(&nm_i->build_lock);
1777 spin_lock_init(&nm_i->free_nid_list_lock);
1778 rwlock_init(&nm_i->nat_tree_lock);
1779
e05df3b1 1780 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
79b5793b 1781 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
e05df3b1
JK		 1782 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
1783 if (!version_bitmap)
1784 return -EFAULT;
1785
79b5793b
AG		 1786 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
1787 GFP_KERNEL);
1788 if (!nm_i->nat_bitmap)
1789 return -ENOMEM;
e05df3b1
JK		 1790 return 0;
1791}
1792
1793int build_node_manager(struct f2fs_sb_info *sbi)
1794{
1795 int err;
1796
1797 sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL);
1798 if (!sbi->nm_info)
1799 return -ENOMEM;
1800
1801 err = init_node_manager(sbi);
1802 if (err)
1803 return err;
1804
1805 build_free_nids(sbi);
1806 return 0;
1807}
1808
1809void destroy_node_manager(struct f2fs_sb_info *sbi)
1810{
1811 struct f2fs_nm_info *nm_i = NM_I(sbi);
1812 struct free_nid *i, *next_i;
1813 struct nat_entry *natvec[NATVEC_SIZE];
1814 nid_t nid = 0;
1815 unsigned int found;
1816
1817 if (!nm_i)
1818 return;
1819
1820 /* destroy free nid list */
1821 spin_lock(&nm_i->free_nid_list_lock);
1822 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
5d56b671 1823 f2fs_bug_on(i->state == NID_ALLOC);
e05df3b1
JK		 1824 __del_from_free_nid_list(i);
1825 nm_i->fcnt--;
1826 }
5d56b671 1827 f2fs_bug_on(nm_i->fcnt);
e05df3b1
JK		 1828 spin_unlock(&nm_i->free_nid_list_lock);
1829
1830 /* destroy nat cache */
1831 write_lock(&nm_i->nat_tree_lock);
1832 while ((found = __gang_lookup_nat_cache(nm_i,
1833 nid, NATVEC_SIZE, natvec))) {
1834 unsigned idx;
1835 for (idx = 0; idx < found; idx++) {
1836 struct nat_entry *e = natvec[idx];
1837 nid = nat_get_nid(e) + 1;
1838 __del_from_nat_cache(nm_i, e);
1839 }
1840 }
5d56b671 1841 f2fs_bug_on(nm_i->nat_cnt);
e05df3b1
JK		 1842 write_unlock(&nm_i->nat_tree_lock);
1843
1844 kfree(nm_i->nat_bitmap);
1845 sbi->nm_info = NULL;
1846 kfree(nm_i);
1847}
1848
6e6093a8 1849int __init create_node_manager_caches(void)
e05df3b1
JK		 1850{
1851 nat_entry_slab = f2fs_kmem_cache_create("nat_entry",
1852 sizeof(struct nat_entry), NULL);
1853 if (!nat_entry_slab)
1854 return -ENOMEM;
1855
1856 free_nid_slab = f2fs_kmem_cache_create("free_nid",
1857 sizeof(struct free_nid), NULL);
1858 if (!free_nid_slab) {
1859 kmem_cache_destroy(nat_entry_slab);
1860 return -ENOMEM;
1861 }
1862 return 0;
1863}
1864
1865void destroy_node_manager_caches(void)
1866{
1867 kmem_cache_destroy(free_nid_slab);
1868 kmem_cache_destroy(nat_entry_slab);
1869}
Linux 6.x block layer and io_uring tree(s)