projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: fix features filename in sysfs documentation
[linux-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"
87905682 22#include "xattr.h"
9e4ded3f 23#include "trace.h"
51dd6249 24#include <trace/events/f2fs.h>
e05df3b1 25
68afcf2d 26#define on_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock)
f978f5a0 27
e05df3b1
JK		 28static struct kmem_cache *nat_entry_slab;
29static struct kmem_cache *free_nid_slab;
aec71382 30static struct kmem_cache *nat_entry_set_slab;
e05df3b1 31
6fb03f3a 32bool available_free_memory(struct f2fs_sb_info *sbi, int type)
cdfc41c1 33{
6fb03f3a 34 struct f2fs_nm_info *nm_i = NM_I(sbi);
cdfc41c1 35 struct sysinfo val;
e5e7ea3c 36 unsigned long avail_ram;
cdfc41c1 37 unsigned long mem_size = 0;
6fb03f3a 38 bool res = false;
cdfc41c1
JK		 39
40 si_meminfo(&val);
e5e7ea3c
JK		 41
42 /* only uses low memory */
43 avail_ram = val.totalram - val.totalhigh;
44
429511cd
CY		 45 /*
46 * give 25%, 25%, 50%, 50%, 50% memory for each components respectively
47 */
6fb03f3a 48 if (type == FREE_NIDS) {
9a4ffdf5 49 mem_size = (nm_i->nid_cnt[FREE_NID] *
b8559dc2 50 sizeof(struct free_nid)) >> PAGE_SHIFT;
e5e7ea3c 51 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
6fb03f3a 52 } else if (type == NAT_ENTRIES) {
e5e7ea3c 53 mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >>
09cbfeaf 54 PAGE_SHIFT;
e5e7ea3c 55 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
e589c2c4
JK		 56 if (excess_cached_nats(sbi))
57 res = false;
a1257023
JK		 58 } else if (type == DIRTY_DENTS) {
59 if (sbi->sb->s_bdi->wb.dirty_exceeded)
60 return false;
61 mem_size = get_pages(sbi, F2FS_DIRTY_DENTS);
62 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
e5e7ea3c
JK		 63 } else if (type == INO_ENTRIES) {
64 int i;
65
39d787be 66 for (i = 0; i < MAX_INO_ENTRY; i++)
8f73cbb7
KM		 67 mem_size += sbi->im[i].ino_num *
68 sizeof(struct ino_entry);
69 mem_size >>= PAGE_SHIFT;
e5e7ea3c 70 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
429511cd 71 } else if (type == EXTENT_CACHE) {
7441ccef
JK		 72 mem_size = (atomic_read(&sbi->total_ext_tree) *
73 sizeof(struct extent_tree) +
429511cd 74 atomic_read(&sbi->total_ext_node) *
09cbfeaf 75 sizeof(struct extent_node)) >> PAGE_SHIFT;
429511cd 76 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
57864ae5
JK		 77 } else if (type == INMEM_PAGES) {
78 /* it allows 20% / total_ram for inmemory pages */
79 mem_size = get_pages(sbi, F2FS_INMEM_PAGES);
80 res = mem_size < (val.totalram / 5);
1e84371f 81 } else {
1663cae4
JK		 82 if (!sbi->sb->s_bdi->wb.dirty_exceeded)
83 return true;
6fb03f3a
JK		 84 }
85 return res;
cdfc41c1
JK		 86}
87
e05df3b1
JK		 88static void clear_node_page_dirty(struct page *page)
89{
90 struct address_space *mapping = page->mapping;
e05df3b1
JK		 91 unsigned int long flags;
92
93 if (PageDirty(page)) {
b93b0163
MW		 94 xa_lock_irqsave(&mapping->i_pages, flags);
95 radix_tree_tag_clear(&mapping->i_pages,
e05df3b1
JK		 96 page_index(page),
97 PAGECACHE_TAG_DIRTY);
b93b0163 98 xa_unlock_irqrestore(&mapping->i_pages, flags);
e05df3b1
JK		 99
100 clear_page_dirty_for_io(page);
4081363f 101 dec_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES);
e05df3b1
JK		 102 }
103 ClearPageUptodate(page);
104}
105
106static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
107{
108 pgoff_t index = current_nat_addr(sbi, nid);
109 return get_meta_page(sbi, index);
110}
111
112static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
113{
114 struct page *src_page;
115 struct page *dst_page;
116 pgoff_t src_off;
117 pgoff_t dst_off;
118 void *src_addr;
119 void *dst_addr;
120 struct f2fs_nm_info *nm_i = NM_I(sbi);
121
122 src_off = current_nat_addr(sbi, nid);
123 dst_off = next_nat_addr(sbi, src_off);
124
125 /* get current nat block page with lock */
126 src_page = get_meta_page(sbi, src_off);
e05df3b1 127 dst_page = grab_meta_page(sbi, dst_off);
9850cf4a 128 f2fs_bug_on(sbi, PageDirty(src_page));
e05df3b1
JK		 129
130 src_addr = page_address(src_page);
131 dst_addr = page_address(dst_page);
09cbfeaf 132 memcpy(dst_addr, src_addr, PAGE_SIZE);
e05df3b1
JK		 133 set_page_dirty(dst_page);
134 f2fs_put_page(src_page, 1);
135
136 set_to_next_nat(nm_i, nid);
137
138 return dst_page;
139}
140
12f9ef37
YH		 141static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail)
142{
143 struct nat_entry *new;
144
145 if (no_fail)
2882d343 146 new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37 147 else
2882d343 148 new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37
YH		 149 if (new) {
150 nat_set_nid(new, nid);
151 nat_reset_flag(new);
152 }
153 return new;
154}
155
156static void __free_nat_entry(struct nat_entry *e)
157{
158 kmem_cache_free(nat_entry_slab, e);
159}
160
161/* must be locked by nat_tree_lock */
162static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
163 struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail)
164{
165 if (no_fail)
166 f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne);
167 else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne))
168 return NULL;
169
170 if (raw_ne)
171 node_info_from_raw_nat(&ne->ni, raw_ne);
172 list_add_tail(&ne->list, &nm_i->nat_entries);
173 nm_i->nat_cnt++;
174 return ne;
175}
176
e05df3b1
JK		 177static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
178{
179 return radix_tree_lookup(&nm_i->nat_root, n);
180}
181
182static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
183 nid_t start, unsigned int nr, struct nat_entry **ep)
184{
185 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
186}
187
188static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
189{
190 list_del(&e->list);
191 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
192 nm_i->nat_cnt--;
12f9ef37 193 __free_nat_entry(e);
e05df3b1
JK		 194}
195
780de47c
CY		 196static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
197 struct nat_entry *ne)
309cc2b6
JK		 198{
199 nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid);
200 struct nat_entry_set *head;
201
309cc2b6
JK		 202 head = radix_tree_lookup(&nm_i->nat_set_root, set);
203 if (!head) {
80c54505 204 head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS);
309cc2b6
JK		 205
206 INIT_LIST_HEAD(&head->entry_list);
207 INIT_LIST_HEAD(&head->set_list);
208 head->set = set;
209 head->entry_cnt = 0;
9be32d72 210 f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
309cc2b6 211 }
780de47c
CY		 212 return head;
213}
214
215static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
216 struct nat_entry *ne)
217{
218 struct nat_entry_set *head;
219 bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR;
220
221 if (!new_ne)
222 head = __grab_nat_entry_set(nm_i, ne);
223
224 /*
225 * update entry_cnt in below condition:
226 * 1. update NEW_ADDR to valid block address;
227 * 2. update old block address to new one;
228 */
229 if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
230 !get_nat_flag(ne, IS_DIRTY)))
231 head->entry_cnt++;
232
233 set_nat_flag(ne, IS_PREALLOC, new_ne);
febeca6d
CY		 234
235 if (get_nat_flag(ne, IS_DIRTY))
236 goto refresh_list;
237
309cc2b6 238 nm_i->dirty_nat_cnt++;
309cc2b6 239 set_nat_flag(ne, IS_DIRTY, true);
febeca6d 240refresh_list:
780de47c 241 if (new_ne)
febeca6d
CY		 242 list_del_init(&ne->list);
243 else
244 list_move_tail(&ne->list, &head->entry_list);
309cc2b6
JK		 245}
246
247static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
0b28b71e 248 struct nat_entry_set *set, struct nat_entry *ne)
309cc2b6 249{
0b28b71e
KM		 250 list_move_tail(&ne->list, &nm_i->nat_entries);
251 set_nat_flag(ne, IS_DIRTY, false);
252 set->entry_cnt--;
253 nm_i->dirty_nat_cnt--;
309cc2b6
JK		 254}
255
256static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
257 nid_t start, unsigned int nr, struct nat_entry_set **ep)
258{
259 return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep,
260 start, nr);
261}
262
2dcf51ab 263int need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK		 264{
265 struct f2fs_nm_info *nm_i = NM_I(sbi);
266 struct nat_entry *e;
2dcf51ab 267 bool need = false;
e05df3b1 268
b873b798 269 down_read(&nm_i->nat_tree_lock);
e05df3b1 270 e = __lookup_nat_cache(nm_i, nid);
2dcf51ab
JK		 271 if (e) {
272 if (!get_nat_flag(e, IS_CHECKPOINTED) &&
273 !get_nat_flag(e, HAS_FSYNCED_INODE))
274 need = true;
275 }
b873b798 276 up_read(&nm_i->nat_tree_lock);
2dcf51ab 277 return need;
e05df3b1
JK		 278}
279
2dcf51ab 280bool is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
479f40c4
JK		 281{
282 struct f2fs_nm_info *nm_i = NM_I(sbi);
283 struct nat_entry *e;
2dcf51ab 284 bool is_cp = true;
479f40c4 285
b873b798 286 down_read(&nm_i->nat_tree_lock);
2dcf51ab
JK		 287 e = __lookup_nat_cache(nm_i, nid);
288 if (e && !get_nat_flag(e, IS_CHECKPOINTED))
289 is_cp = false;
b873b798 290 up_read(&nm_i->nat_tree_lock);
2dcf51ab 291 return is_cp;
479f40c4
JK		 292}
293
88bd02c9 294bool need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
b6fe5873
JK		 295{
296 struct f2fs_nm_info *nm_i = NM_I(sbi);
297 struct nat_entry *e;
88bd02c9 298 bool need_update = true;
b6fe5873 299
b873b798 300 down_read(&nm_i->nat_tree_lock);
88bd02c9
JK		 301 e = __lookup_nat_cache(nm_i, ino);
302 if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
303 (get_nat_flag(e, IS_CHECKPOINTED) ||
304 get_nat_flag(e, HAS_FSYNCED_INODE)))
305 need_update = false;
b873b798 306 up_read(&nm_i->nat_tree_lock);
88bd02c9 307 return need_update;
b6fe5873
JK		 308}
309
12f9ef37 310/* must be locked by nat_tree_lock */
1515aef0 311static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
e05df3b1
JK		 312 struct f2fs_nat_entry *ne)
313{
1515aef0 314 struct f2fs_nm_info *nm_i = NM_I(sbi);
12f9ef37 315 struct nat_entry *new, *e;
9be32d72 316
12f9ef37
YH		 317 new = __alloc_nat_entry(nid, false);
318 if (!new)
319 return;
320
321 down_write(&nm_i->nat_tree_lock);
e05df3b1 322 e = __lookup_nat_cache(nm_i, nid);
12f9ef37
YH		 323 if (!e)
324 e = __init_nat_entry(nm_i, new, ne, false);
325 else
0c0b471e
EB		 326 f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) ||
327 nat_get_blkaddr(e) !=
328 le32_to_cpu(ne->block_addr) ||
1515aef0 329 nat_get_version(e) != ne->version);
12f9ef37
YH		 330 up_write(&nm_i->nat_tree_lock);
331 if (e != new)
332 __free_nat_entry(new);
e05df3b1
JK		 333}
334
335static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
479f40c4 336 block_t new_blkaddr, bool fsync_done)
e05df3b1
JK		 337{
338 struct f2fs_nm_info *nm_i = NM_I(sbi);
339 struct nat_entry *e;
12f9ef37 340 struct nat_entry *new = __alloc_nat_entry(ni->nid, true);
9be32d72 341
b873b798 342 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 343 e = __lookup_nat_cache(nm_i, ni->nid);
344 if (!e) {
12f9ef37 345 e = __init_nat_entry(nm_i, new, NULL, true);
5c27f4ee 346 copy_node_info(&e->ni, ni);
9850cf4a 347 f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR);
e05df3b1
JK		 348 } else if (new_blkaddr == NEW_ADDR) {
349 /*
350 * when nid is reallocated,
351 * previous nat entry can be remained in nat cache.
352 * So, reinitialize it with new information.
353 */
5c27f4ee 354 copy_node_info(&e->ni, ni);
9850cf4a 355 f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR);
e05df3b1 356 }
12f9ef37
YH		 357 /* let's free early to reduce memory consumption */
358 if (e != new)
359 __free_nat_entry(new);
e05df3b1 360
e05df3b1 361 /* sanity check */
9850cf4a
JK		 362 f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr);
363 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR &&
e05df3b1 364 new_blkaddr == NULL_ADDR);
9850cf4a 365 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR &&
e05df3b1 366 new_blkaddr == NEW_ADDR);
9850cf4a 367 f2fs_bug_on(sbi, nat_get_blkaddr(e) != NEW_ADDR &&
e05df3b1
JK		 368 nat_get_blkaddr(e) != NULL_ADDR &&
369 new_blkaddr == NEW_ADDR);
370
e1c42045 371 /* increment version no as node is removed */
e05df3b1
JK		 372 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
373 unsigned char version = nat_get_version(e);
374 nat_set_version(e, inc_node_version(version));
375 }
376
377 /* change address */
378 nat_set_blkaddr(e, new_blkaddr);
88bd02c9
JK		 379 if (new_blkaddr == NEW_ADDR || new_blkaddr == NULL_ADDR)
380 set_nat_flag(e, IS_CHECKPOINTED, false);
e05df3b1 381 __set_nat_cache_dirty(nm_i, e);
479f40c4
JK		 382
383 /* update fsync_mark if its inode nat entry is still alive */
d5b692b7
CY		 384 if (ni->nid != ni->ino)
385 e = __lookup_nat_cache(nm_i, ni->ino);
88bd02c9
JK		 386 if (e) {
387 if (fsync_done && ni->nid == ni->ino)
388 set_nat_flag(e, HAS_FSYNCED_INODE, true);
389 set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
390 }
b873b798 391 up_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 392}
393
4660f9c0 394int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
e05df3b1
JK		 395{
396 struct f2fs_nm_info *nm_i = NM_I(sbi);
1b38dc8e 397 int nr = nr_shrink;
e05df3b1 398
b873b798
JK		 399 if (!down_write_trylock(&nm_i->nat_tree_lock))
400 return 0;
e05df3b1 401
e05df3b1
JK		 402 while (nr_shrink && !list_empty(&nm_i->nat_entries)) {
403 struct nat_entry *ne;
404 ne = list_first_entry(&nm_i->nat_entries,
405 struct nat_entry, list);
406 __del_from_nat_cache(nm_i, ne);
407 nr_shrink--;
408 }
b873b798 409 up_write(&nm_i->nat_tree_lock);
1b38dc8e 410 return nr - nr_shrink;
e05df3b1
JK		 411}
412
0a8165d7 413/*
e1c42045 414 * This function always returns success
e05df3b1
JK		 415 */
416void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
417{
418 struct f2fs_nm_info *nm_i = NM_I(sbi);
419 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 420 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 421 nid_t start_nid = START_NID(nid);
422 struct f2fs_nat_block *nat_blk;
423 struct page *page = NULL;
424 struct f2fs_nat_entry ne;
425 struct nat_entry *e;
66a82d1f 426 pgoff_t index;
e05df3b1
JK		 427 int i;
428
429 ni->nid = nid;
430
431 /* Check nat cache */
b873b798 432 down_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 433 e = __lookup_nat_cache(nm_i, nid);
434 if (e) {
435 ni->ino = nat_get_ino(e);
436 ni->blk_addr = nat_get_blkaddr(e);
437 ni->version = nat_get_version(e);
b873b798 438 up_read(&nm_i->nat_tree_lock);
e05df3b1 439 return;
1515aef0 440 }
e05df3b1 441
3547ea96
JK		 442 memset(&ne, 0, sizeof(struct f2fs_nat_entry));
443
e05df3b1 444 /* Check current segment summary */
b7ad7512 445 down_read(&curseg->journal_rwsem);
dfc08a12 446 i = lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0);
e05df3b1 447 if (i >= 0) {
dfc08a12 448 ne = nat_in_journal(journal, i);
e05df3b1
JK		 449 node_info_from_raw_nat(ni, &ne);
450 }
b7ad7512 451 up_read(&curseg->journal_rwsem);
66a82d1f
YH		 452 if (i >= 0) {
453 up_read(&nm_i->nat_tree_lock);
e05df3b1 454 goto cache;
66a82d1f 455 }
e05df3b1
JK		 456
457 /* Fill node_info from nat page */
66a82d1f
YH		 458 index = current_nat_addr(sbi, nid);
459 up_read(&nm_i->nat_tree_lock);
460
461 page = get_meta_page(sbi, index);
e05df3b1
JK		 462 nat_blk = (struct f2fs_nat_block *)page_address(page);
463 ne = nat_blk->entries[nid - start_nid];
464 node_info_from_raw_nat(ni, &ne);
465 f2fs_put_page(page, 1);
466cache:
467 /* cache nat entry */
1515aef0 468 cache_nat_entry(sbi, nid, &ne);
e05df3b1
JK		 469}
470
79344efb
JK		 471/*
472 * readahead MAX_RA_NODE number of node pages.
473 */
474static void ra_node_pages(struct page *parent, int start, int n)
475{
476 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
477 struct blk_plug plug;
478 int i, end;
479 nid_t nid;
480
481 blk_start_plug(&plug);
482
483 /* Then, try readahead for siblings of the desired node */
484 end = start + n;
485 end = min(end, NIDS_PER_BLOCK);
486 for (i = start; i < end; i++) {
487 nid = get_nid(parent, i, false);
488 ra_node_page(sbi, nid);
489 }
490
491 blk_finish_plug(&plug);
492}
493
3cf45747
CY		 494pgoff_t get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs)
495{
496 const long direct_index = ADDRS_PER_INODE(dn->inode);
497 const long direct_blks = ADDRS_PER_BLOCK;
498 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
499 unsigned int skipped_unit = ADDRS_PER_BLOCK;
500 int cur_level = dn->cur_level;
501 int max_level = dn->max_level;
502 pgoff_t base = 0;
503
504 if (!dn->max_level)
505 return pgofs + 1;
506
507 while (max_level-- > cur_level)
508 skipped_unit *= NIDS_PER_BLOCK;
509
510 switch (dn->max_level) {
511 case 3:
512 base += 2 * indirect_blks;
513 case 2:
514 base += 2 * direct_blks;
515 case 1:
516 base += direct_index;
517 break;
518 default:
519 f2fs_bug_on(F2FS_I_SB(dn->inode), 1);
520 }
521
522 return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base;
523}
524
0a8165d7 525/*
e05df3b1
JK		 526 * The maximum depth is four.
527 * Offset[0] will have raw inode offset.
528 */
81ca7350 529static int get_node_path(struct inode *inode, long block,
de93653f 530 int offset[4], unsigned int noffset[4])
e05df3b1 531{
81ca7350 532 const long direct_index = ADDRS_PER_INODE(inode);
e05df3b1
JK		 533 const long direct_blks = ADDRS_PER_BLOCK;
534 const long dptrs_per_blk = NIDS_PER_BLOCK;
535 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
536 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
537 int n = 0;
538 int level = 0;
539
540 noffset[0] = 0;
541
542 if (block < direct_index) {
25c0a6e5 543 offset[n] = block;
e05df3b1
JK		 544 goto got;
545 }
546 block -= direct_index;
547 if (block < direct_blks) {
548 offset[n++] = NODE_DIR1_BLOCK;
549 noffset[n] = 1;
25c0a6e5 550 offset[n] = block;
e05df3b1
JK		 551 level = 1;
552 goto got;
553 }
554 block -= direct_blks;
555 if (block < direct_blks) {
556 offset[n++] = NODE_DIR2_BLOCK;
557 noffset[n] = 2;
25c0a6e5 558 offset[n] = block;
e05df3b1
JK		 559 level = 1;
560 goto got;
561 }
562 block -= direct_blks;
563 if (block < indirect_blks) {
564 offset[n++] = NODE_IND1_BLOCK;
565 noffset[n] = 3;
566 offset[n++] = block / direct_blks;
567 noffset[n] = 4 + offset[n - 1];
25c0a6e5 568 offset[n] = block % direct_blks;
e05df3b1
JK		 569 level = 2;
570 goto got;
571 }
572 block -= indirect_blks;
573 if (block < indirect_blks) {
574 offset[n++] = NODE_IND2_BLOCK;
575 noffset[n] = 4 + dptrs_per_blk;
576 offset[n++] = block / direct_blks;
577 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
25c0a6e5 578 offset[n] = block % direct_blks;
e05df3b1
JK		 579 level = 2;
580 goto got;
581 }
582 block -= indirect_blks;
583 if (block < dindirect_blks) {
584 offset[n++] = NODE_DIND_BLOCK;
585 noffset[n] = 5 + (dptrs_per_blk * 2);
586 offset[n++] = block / indirect_blks;
587 noffset[n] = 6 + (dptrs_per_blk * 2) +
588 offset[n - 1] * (dptrs_per_blk + 1);
589 offset[n++] = (block / direct_blks) % dptrs_per_blk;
590 noffset[n] = 7 + (dptrs_per_blk * 2) +
591 offset[n - 2] * (dptrs_per_blk + 1) +
592 offset[n - 1];
25c0a6e5 593 offset[n] = block % direct_blks;
e05df3b1
JK		 594 level = 3;
595 goto got;
596 } else {
adb6dc19 597 return -E2BIG;
e05df3b1
JK		 598 }
599got:
600 return level;
601}
602
603/*
604 * Caller should call f2fs_put_dnode(dn).
4f4124d0
CY		 605 * Also, it should grab and release a rwsem by calling f2fs_lock_op() and
606 * f2fs_unlock_op() only if ro is not set RDONLY_NODE.
39936837 607 * In the case of RDONLY_NODE, we don't need to care about mutex.
e05df3b1 608 */
266e97a8 609int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
e05df3b1 610{
4081363f 611 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 612 struct page *npage[4];
f1a3b98e 613 struct page *parent = NULL;
e05df3b1
JK		 614 int offset[4];
615 unsigned int noffset[4];
616 nid_t nids[4];
3cf45747 617 int level, i = 0;
e05df3b1
JK		 618 int err = 0;
619
81ca7350 620 level = get_node_path(dn->inode, index, offset, noffset);
adb6dc19
JK		 621 if (level < 0)
622 return level;
e05df3b1
JK		 623
624 nids[0] = dn->inode->i_ino;
1646cfac 625 npage[0] = dn->inode_page;
e05df3b1 626
1646cfac
JK		 627 if (!npage[0]) {
628 npage[0] = get_node_page(sbi, nids[0]);
629 if (IS_ERR(npage[0]))
630 return PTR_ERR(npage[0]);
631 }
f1a3b98e
JK		 632
633 /* if inline_data is set, should not report any block indices */
634 if (f2fs_has_inline_data(dn->inode) && index) {
76629165 635 err = -ENOENT;
f1a3b98e
JK		 636 f2fs_put_page(npage[0], 1);
637 goto release_out;
638 }
639
e05df3b1 640 parent = npage[0];
52c2db3f
CL		 641 if (level != 0)
642 nids[1] = get_nid(parent, offset[0], true);
e05df3b1
JK		 643 dn->inode_page = npage[0];
644 dn->inode_page_locked = true;
645
646 /* get indirect or direct nodes */
647 for (i = 1; i <= level; i++) {
648 bool done = false;
649
266e97a8 650 if (!nids[i] && mode == ALLOC_NODE) {
e05df3b1
JK		 651 /* alloc new node */
652 if (!alloc_nid(sbi, &(nids[i]))) {
e05df3b1
JK		 653 err = -ENOSPC;
654 goto release_pages;
655 }
656
657 dn->nid = nids[i];
5f4ce6ab 658 npage[i] = new_node_page(dn, noffset[i]);
e05df3b1
JK		 659 if (IS_ERR(npage[i])) {
660 alloc_nid_failed(sbi, nids[i]);
e05df3b1
JK		 661 err = PTR_ERR(npage[i]);
662 goto release_pages;
663 }
664
665 set_nid(parent, offset[i - 1], nids[i], i == 1);
666 alloc_nid_done(sbi, nids[i]);
e05df3b1 667 done = true;
266e97a8 668 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
e05df3b1
JK		 669 npage[i] = get_node_page_ra(parent, offset[i - 1]);
670 if (IS_ERR(npage[i])) {
671 err = PTR_ERR(npage[i]);
672 goto release_pages;
673 }
674 done = true;
675 }
676 if (i == 1) {
677 dn->inode_page_locked = false;
678 unlock_page(parent);
679 } else {
680 f2fs_put_page(parent, 1);
681 }
682
683 if (!done) {
684 npage[i] = get_node_page(sbi, nids[i]);
685 if (IS_ERR(npage[i])) {
686 err = PTR_ERR(npage[i]);
687 f2fs_put_page(npage[0], 0);
688 goto release_out;
689 }
690 }
691 if (i < level) {
692 parent = npage[i];
693 nids[i + 1] = get_nid(parent, offset[i], false);
694 }
695 }
696 dn->nid = nids[level];
697 dn->ofs_in_node = offset[level];
698 dn->node_page = npage[level];
7a2af766
CY		 699 dn->data_blkaddr = datablock_addr(dn->inode,
700 dn->node_page, dn->ofs_in_node);
e05df3b1
JK		 701 return 0;
702
703release_pages:
704 f2fs_put_page(parent, 1);
705 if (i > 1)
706 f2fs_put_page(npage[0], 0);
707release_out:
708 dn->inode_page = NULL;
709 dn->node_page = NULL;
3cf45747
CY		 710 if (err == -ENOENT) {
711 dn->cur_level = i;
712 dn->max_level = level;
0a2aa8fb 713 dn->ofs_in_node = offset[level];
3cf45747 714 }
e05df3b1
JK		 715 return err;
716}
717
718static void truncate_node(struct dnode_of_data *dn)
719{
4081363f 720 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1
JK		 721 struct node_info ni;
722
723 get_node_info(sbi, dn->nid, &ni);
e05df3b1 724
e05df3b1 725 /* Deallocate node address */
71e9fec5 726 invalidate_blocks(sbi, ni.blk_addr);
000519f2 727 dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino);
479f40c4 728 set_node_addr(sbi, &ni, NULL_ADDR, false);
e05df3b1
JK		 729
730 if (dn->nid == dn->inode->i_ino) {
731 remove_orphan_inode(sbi, dn->nid);
732 dec_valid_inode_count(sbi);
0f18b462 733 f2fs_inode_synced(dn->inode);
e05df3b1 734 }
000519f2 735
e05df3b1 736 clear_node_page_dirty(dn->node_page);
caf0047e 737 set_sbi_flag(sbi, SBI_IS_DIRTY);
e05df3b1
JK		 738
739 f2fs_put_page(dn->node_page, 1);
bf39c00a
JK		 740
741 invalidate_mapping_pages(NODE_MAPPING(sbi),
742 dn->node_page->index, dn->node_page->index);
743
e05df3b1 744 dn->node_page = NULL;
51dd6249 745 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
e05df3b1
JK		 746}
747
748static int truncate_dnode(struct dnode_of_data *dn)
749{
e05df3b1
JK		 750 struct page *page;
751
752 if (dn->nid == 0)
753 return 1;
754
755 /* get direct node */
4081363f 756 page = get_node_page(F2FS_I_SB(dn->inode), dn->nid);
e05df3b1
JK		 757 if (IS_ERR(page) && PTR_ERR(page) == -ENOENT)
758 return 1;
759 else if (IS_ERR(page))
760 return PTR_ERR(page);
761
762 /* Make dnode_of_data for parameter */
763 dn->node_page = page;
764 dn->ofs_in_node = 0;
765 truncate_data_blocks(dn);
766 truncate_node(dn);
767 return 1;
768}
769
770static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
771 int ofs, int depth)
772{
e05df3b1
JK		 773 struct dnode_of_data rdn = *dn;
774 struct page *page;
775 struct f2fs_node *rn;
776 nid_t child_nid;
777 unsigned int child_nofs;
778 int freed = 0;
779 int i, ret;
780
781 if (dn->nid == 0)
782 return NIDS_PER_BLOCK + 1;
783
51dd6249
NJ		 784 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
785
4081363f 786 page = get_node_page(F2FS_I_SB(dn->inode), dn->nid);
51dd6249
NJ		 787 if (IS_ERR(page)) {
788 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
e05df3b1 789 return PTR_ERR(page);
51dd6249 790 }
e05df3b1 791
79344efb
JK		 792 ra_node_pages(page, ofs, NIDS_PER_BLOCK);
793
45590710 794 rn = F2FS_NODE(page);
e05df3b1
JK		 795 if (depth < 3) {
796 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
797 child_nid = le32_to_cpu(rn->in.nid[i]);
798 if (child_nid == 0)
799 continue;
800 rdn.nid = child_nid;
801 ret = truncate_dnode(&rdn);
802 if (ret < 0)
803 goto out_err;
12719ae1
JK		 804 if (set_nid(page, i, 0, false))
805 dn->node_changed = true;
e05df3b1
JK		 806 }
807 } else {
808 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
809 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
810 child_nid = le32_to_cpu(rn->in.nid[i]);
811 if (child_nid == 0) {
812 child_nofs += NIDS_PER_BLOCK + 1;
813 continue;
814 }
815 rdn.nid = child_nid;
816 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
817 if (ret == (NIDS_PER_BLOCK + 1)) {
12719ae1
JK		 818 if (set_nid(page, i, 0, false))
819 dn->node_changed = true;
e05df3b1
JK		 820 child_nofs += ret;
821 } else if (ret < 0 && ret != -ENOENT) {
822 goto out_err;
823 }
824 }
825 freed = child_nofs;
826 }
827
828 if (!ofs) {
829 /* remove current indirect node */
830 dn->node_page = page;
831 truncate_node(dn);
832 freed++;
833 } else {
834 f2fs_put_page(page, 1);
835 }
51dd6249 836 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
e05df3b1
JK		 837 return freed;
838
839out_err:
840 f2fs_put_page(page, 1);
51dd6249 841 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
e05df3b1
JK		 842 return ret;
843}
844
845static int truncate_partial_nodes(struct dnode_of_data *dn,
846 struct f2fs_inode *ri, int *offset, int depth)
847{
e05df3b1
JK		 848 struct page *pages[2];
849 nid_t nid[3];
850 nid_t child_nid;
851 int err = 0;
852 int i;
853 int idx = depth - 2;
854
855 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
856 if (!nid[0])
857 return 0;
858
859 /* get indirect nodes in the path */
a225dca3 860 for (i = 0; i < idx + 1; i++) {
e1c42045 861 /* reference count'll be increased */
4081363f 862 pages[i] = get_node_page(F2FS_I_SB(dn->inode), nid[i]);
e05df3b1 863 if (IS_ERR(pages[i])) {
e05df3b1 864 err = PTR_ERR(pages[i]);
a225dca3 865 idx = i - 1;
e05df3b1
JK		 866 goto fail;
867 }
868 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
869 }
870
79344efb
JK		 871 ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK);
872
e05df3b1 873 /* free direct nodes linked to a partial indirect node */
a225dca3 874 for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) {
e05df3b1
JK		 875 child_nid = get_nid(pages[idx], i, false);
876 if (!child_nid)
877 continue;
878 dn->nid = child_nid;
879 err = truncate_dnode(dn);
880 if (err < 0)
881 goto fail;
12719ae1
JK		 882 if (set_nid(pages[idx], i, 0, false))
883 dn->node_changed = true;
e05df3b1
JK		 884 }
885
a225dca3 886 if (offset[idx + 1] == 0) {
e05df3b1
JK		 887 dn->node_page = pages[idx];
888 dn->nid = nid[idx];
889 truncate_node(dn);
890 } else {
891 f2fs_put_page(pages[idx], 1);
892 }
893 offset[idx]++;
a225dca3 894 offset[idx + 1] = 0;
895 idx--;
e05df3b1 896fail:
a225dca3 897 for (i = idx; i >= 0; i--)
e05df3b1 898 f2fs_put_page(pages[i], 1);
51dd6249
NJ		 899
900 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
901
e05df3b1
JK		 902 return err;
903}
904
0a8165d7 905/*
e05df3b1
JK		 906 * All the block addresses of data and nodes should be nullified.
907 */
908int truncate_inode_blocks(struct inode *inode, pgoff_t from)
909{
4081363f 910 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e05df3b1
JK		 911 int err = 0, cont = 1;
912 int level, offset[4], noffset[4];
7dd690c8 913 unsigned int nofs = 0;
58bfaf44 914 struct f2fs_inode *ri;
e05df3b1
JK		 915 struct dnode_of_data dn;
916 struct page *page;
917
51dd6249
NJ		 918 trace_f2fs_truncate_inode_blocks_enter(inode, from);
919
81ca7350 920 level = get_node_path(inode, from, offset, noffset);
adb6dc19
JK		 921 if (level < 0)
922 return level;
ff373558 923
e05df3b1 924 page = get_node_page(sbi, inode->i_ino);
51dd6249
NJ		 925 if (IS_ERR(page)) {
926 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
e05df3b1 927 return PTR_ERR(page);
51dd6249 928 }
e05df3b1
JK		 929
930 set_new_dnode(&dn, inode, page, NULL, 0);
931 unlock_page(page);
932
58bfaf44 933 ri = F2FS_INODE(page);
e05df3b1
JK		 934 switch (level) {
935 case 0:
936 case 1:
937 nofs = noffset[1];
938 break;
939 case 2:
940 nofs = noffset[1];
941 if (!offset[level - 1])
942 goto skip_partial;
58bfaf44 943 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 944 if (err < 0 && err != -ENOENT)
945 goto fail;
946 nofs += 1 + NIDS_PER_BLOCK;
947 break;
948 case 3:
949 nofs = 5 + 2 * NIDS_PER_BLOCK;
950 if (!offset[level - 1])
951 goto skip_partial;
58bfaf44 952 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 953 if (err < 0 && err != -ENOENT)
954 goto fail;
955 break;
956 default:
957 BUG();
958 }
959
960skip_partial:
961 while (cont) {
58bfaf44 962 dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
e05df3b1
JK		 963 switch (offset[0]) {
964 case NODE_DIR1_BLOCK:
965 case NODE_DIR2_BLOCK:
966 err = truncate_dnode(&dn);
967 break;
968
969 case NODE_IND1_BLOCK:
970 case NODE_IND2_BLOCK:
971 err = truncate_nodes(&dn, nofs, offset[1], 2);
972 break;
973
974 case NODE_DIND_BLOCK:
975 err = truncate_nodes(&dn, nofs, offset[1], 3);
976 cont = 0;
977 break;
978
979 default:
980 BUG();
981 }
982 if (err < 0 && err != -ENOENT)
983 goto fail;
984 if (offset[1] == 0 &&
58bfaf44 985 ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) {
e05df3b1 986 lock_page(page);
ff373558 987 BUG_ON(page->mapping != NODE_MAPPING(sbi));
fec1d657 988 f2fs_wait_on_page_writeback(page, NODE, true);
58bfaf44 989 ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
e05df3b1
JK		 990 set_page_dirty(page);
991 unlock_page(page);
992 }
993 offset[1] = 0;
994 offset[0]++;
995 nofs += err;
996 }
997fail:
998 f2fs_put_page(page, 0);
51dd6249 999 trace_f2fs_truncate_inode_blocks_exit(inode, err);
e05df3b1
JK		 1000 return err > 0 ? 0 : err;
1001}
1002
9c77f754
JK		 1003/* caller must lock inode page */
1004int truncate_xattr_node(struct inode *inode)
4f16fb0f 1005{
4081363f 1006 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4f16fb0f
JK		 1007 nid_t nid = F2FS_I(inode)->i_xattr_nid;
1008 struct dnode_of_data dn;
1009 struct page *npage;
1010
1011 if (!nid)
1012 return 0;
1013
1014 npage = get_node_page(sbi, nid);
1015 if (IS_ERR(npage))
1016 return PTR_ERR(npage);
1017
205b9822 1018 f2fs_i_xnid_write(inode, 0);
65985d93 1019
9c77f754 1020 set_new_dnode(&dn, inode, NULL, npage, nid);
4f16fb0f
JK		 1021 truncate_node(&dn);
1022 return 0;
1023}
1024
39936837 1025/*
4f4124d0
CY		 1026 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
1027 * f2fs_unlock_op().
39936837 1028 */
13ec7297 1029int remove_inode_page(struct inode *inode)
e05df3b1 1030{
e05df3b1 1031 struct dnode_of_data dn;
13ec7297 1032 int err;
e05df3b1 1033
c2e69583 1034 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
13ec7297
CY		 1035 err = get_dnode_of_data(&dn, 0, LOOKUP_NODE);
1036 if (err)
1037 return err;
e05df3b1 1038
9c77f754 1039 err = truncate_xattr_node(inode);
13ec7297 1040 if (err) {
c2e69583 1041 f2fs_put_dnode(&dn);
13ec7297 1042 return err;
e05df3b1 1043 }
c2e69583
JK		 1044
1045 /* remove potential inline_data blocks */
1046 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1047 S_ISLNK(inode->i_mode))
1048 truncate_data_blocks_range(&dn, 1);
1049
e1c42045 1050 /* 0 is possible, after f2fs_new_inode() has failed */
9850cf4a 1051 f2fs_bug_on(F2FS_I_SB(inode),
0eb0adad 1052 inode->i_blocks != 0 && inode->i_blocks != 8);
c2e69583
JK		 1053
1054 /* will put inode & node pages */
71e9fec5 1055 truncate_node(&dn);
13ec7297 1056 return 0;
e05df3b1
JK		 1057}
1058
a014e037 1059struct page *new_inode_page(struct inode *inode)
e05df3b1 1060{
e05df3b1
JK		 1061 struct dnode_of_data dn;
1062
1063 /* allocate inode page for new inode */
1064 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
44a83ff6
JK		 1065
1066 /* caller should f2fs_put_page(page, 1); */
5f4ce6ab 1067 return new_node_page(&dn, 0);
e05df3b1
JK		 1068}
1069
5f4ce6ab 1070struct page *new_node_page(struct dnode_of_data *dn, unsigned int ofs)
e05df3b1 1071{
4081363f 1072 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
25cc5d3b 1073 struct node_info new_ni;
e05df3b1
JK		 1074 struct page *page;
1075 int err;
1076
91942321 1077 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
e05df3b1
JK		 1078 return ERR_PTR(-EPERM);
1079
300e129c 1080 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false);
e05df3b1
JK		 1081 if (!page)
1082 return ERR_PTR(-ENOMEM);
1083
0abd675e 1084 if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs))))
9c02740c 1085 goto fail;
0abd675e 1086
25cc5d3b
JK		 1087#ifdef CONFIG_F2FS_CHECK_FS
1088 get_node_info(sbi, dn->nid, &new_ni);
1089 f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR);
1090#endif
1091 new_ni.nid = dn->nid;
e05df3b1 1092 new_ni.ino = dn->inode->i_ino;
25cc5d3b
JK		 1093 new_ni.blk_addr = NULL_ADDR;
1094 new_ni.flag = 0;
1095 new_ni.version = 0;
479f40c4 1096 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
9c02740c 1097
fec1d657 1098 f2fs_wait_on_page_writeback(page, NODE, true);
9c02740c 1099 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
c5667575 1100 set_cold_node(page, S_ISDIR(dn->inode->i_mode));
237c0790
JK		 1101 if (!PageUptodate(page))
1102 SetPageUptodate(page);
12719ae1
JK		 1103 if (set_page_dirty(page))
1104 dn->node_changed = true;
e05df3b1 1105
4bc8e9bc 1106 if (f2fs_has_xattr_block(ofs))
205b9822 1107 f2fs_i_xnid_write(dn->inode, dn->nid);
479bd73a 1108
e05df3b1
JK		 1109 if (ofs == 0)
1110 inc_valid_inode_count(sbi);
e05df3b1
JK		 1111 return page;
1112
1113fail:
71e9fec5 1114 clear_node_page_dirty(page);
e05df3b1
JK		 1115 f2fs_put_page(page, 1);
1116 return ERR_PTR(err);
1117}
1118
56ae674c
JK		 1119/*
1120 * Caller should do after getting the following values.
1121 * 0: f2fs_put_page(page, 0)
86531d6b 1122 * LOCKED_PAGE or error: f2fs_put_page(page, 1)
56ae674c 1123 */
04d328de 1124static int read_node_page(struct page *page, int op_flags)
e05df3b1 1125{
4081363f 1126 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
e05df3b1 1127 struct node_info ni;
cf04e8eb 1128 struct f2fs_io_info fio = {
05ca3632 1129 .sbi = sbi,
cf04e8eb 1130 .type = NODE,
04d328de
MC		 1131 .op = REQ_OP_READ,
1132 .op_flags = op_flags,
05ca3632 1133 .page = page,
4375a336 1134 .encrypted_page = NULL,
cf04e8eb 1135 };
e05df3b1 1136
3bdad3c7
JK		 1137 if (PageUptodate(page))
1138 return LOCKED_PAGE;
1139
e05df3b1
JK		 1140 get_node_info(sbi, page->index, &ni);
1141
6bacf52f 1142 if (unlikely(ni.blk_addr == NULL_ADDR)) {
2bca1e23 1143 ClearPageUptodate(page);
e05df3b1 1144 return -ENOENT;
393ff91f
JK		 1145 }
1146
7a9d7548 1147 fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr;
05ca3632 1148 return f2fs_submit_page_bio(&fio);
e05df3b1
JK		 1149}
1150
0a8165d7 1151/*
e05df3b1
JK		 1152 * Readahead a node page
1153 */
1154void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
1155{
e05df3b1 1156 struct page *apage;
56ae674c 1157 int err;
e05df3b1 1158
e8458725
CY		 1159 if (!nid)
1160 return;
1161 f2fs_bug_on(sbi, check_nid_range(sbi, nid));
1162
999270de 1163 rcu_read_lock();
b93b0163 1164 apage = radix_tree_lookup(&NODE_MAPPING(sbi)->i_pages, nid);
999270de
FL		 1165 rcu_read_unlock();
1166 if (apage)
393ff91f 1167 return;
e05df3b1 1168
300e129c 1169 apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1170 if (!apage)
1171 return;
1172
70246286 1173 err = read_node_page(apage, REQ_RAHEAD);
86531d6b 1174 f2fs_put_page(apage, err ? 1 : 0);
e05df3b1
JK		 1175}
1176
17a0ee55 1177static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid,
0e022ea8 1178 struct page *parent, int start)
e05df3b1 1179{
e05df3b1 1180 struct page *page;
0e022ea8 1181 int err;
e05df3b1 1182
e05df3b1
JK		 1183 if (!nid)
1184 return ERR_PTR(-ENOENT);
e8458725 1185 f2fs_bug_on(sbi, check_nid_range(sbi, nid));
afcb7ca0 1186repeat:
300e129c 1187 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1188 if (!page)
1189 return ERR_PTR(-ENOMEM);
1190
70fd7614 1191 err = read_node_page(page, 0);
86531d6b
JK		 1192 if (err < 0) {
1193 f2fs_put_page(page, 1);
e05df3b1 1194 return ERR_PTR(err);
86531d6b 1195 } else if (err == LOCKED_PAGE) {
1f258ec1 1196 err = 0;
56ae674c 1197 goto page_hit;
86531d6b 1198 }
e05df3b1 1199
0e022ea8 1200 if (parent)
79344efb 1201 ra_node_pages(parent, start + 1, MAX_RA_NODE);
c718379b 1202
e05df3b1 1203 lock_page(page);
0e022ea8 1204
4ef51a8f 1205 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
afcb7ca0
JK		 1206 f2fs_put_page(page, 1);
1207 goto repeat;
1208 }
1563ac75 1209
1f258ec1
CY		 1210 if (unlikely(!PageUptodate(page))) {
1211 err = -EIO;
1563ac75 1212 goto out_err;
1f258ec1 1213 }
704956ec
CY		 1214
1215 if (!f2fs_inode_chksum_verify(sbi, page)) {
1216 err = -EBADMSG;
1217 goto out_err;
1218 }
e0f56cb4 1219page_hit:
0c9df7fb 1220 if(unlikely(nid != nid_of_node(page))) {
1f258ec1
CY		 1221 f2fs_msg(sbi->sb, KERN_WARNING, "inconsistent node block, "
1222 "nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]",
1223 nid, nid_of_node(page), ino_of_node(page),
1224 ofs_of_node(page), cpver_of_node(page),
1225 next_blkaddr_of_node(page));
1f258ec1 1226 err = -EINVAL;
0c9df7fb 1227out_err:
ee605234 1228 ClearPageUptodate(page);
0c9df7fb 1229 f2fs_put_page(page, 1);
1f258ec1 1230 return ERR_PTR(err);
0c9df7fb 1231 }
e05df3b1
JK		 1232 return page;
1233}
1234
0e022ea8
CY		 1235struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
1236{
1237 return __get_node_page(sbi, nid, NULL, 0);
1238}
1239
1240struct page *get_node_page_ra(struct page *parent, int start)
1241{
1242 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
1243 nid_t nid = get_nid(parent, start, false);
1244
1245 return __get_node_page(sbi, nid, parent, start);
1246}
1247
2049d4fc
JK		 1248static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino)
1249{
1250 struct inode *inode;
1251 struct page *page;
0f3311a8 1252 int ret;
2049d4fc
JK		 1253
1254 /* should flush inline_data before evict_inode */
1255 inode = ilookup(sbi->sb, ino);
1256 if (!inode)
1257 return;
1258
01eccef7
CY		 1259 page = f2fs_pagecache_get_page(inode->i_mapping, 0,
1260 FGP_LOCK|FGP_NOWAIT, 0);
2049d4fc
JK		 1261 if (!page)
1262 goto iput_out;
1263
1264 if (!PageUptodate(page))
1265 goto page_out;
1266
1267 if (!PageDirty(page))
1268 goto page_out;
1269
1270 if (!clear_page_dirty_for_io(page))
1271 goto page_out;
1272
0f3311a8
CY		 1273 ret = f2fs_write_inline_data(inode, page);
1274 inode_dec_dirty_pages(inode);
933439c8 1275 remove_dirty_inode(inode);
0f3311a8 1276 if (ret)
2049d4fc
JK		 1277 set_page_dirty(page);
1278page_out:
4a6de50d 1279 f2fs_put_page(page, 1);
2049d4fc
JK		 1280iput_out:
1281 iput(inode);
1282}
1283
608514de
JK		 1284static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
1285{
028a63a6 1286 pgoff_t index;
608514de
JK		 1287 struct pagevec pvec;
1288 struct page *last_page = NULL;
028a63a6 1289 int nr_pages;
608514de 1290
86679820 1291 pagevec_init(&pvec);
608514de 1292 index = 0;
028a63a6
JK		 1293
1294 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1295 PAGECACHE_TAG_DIRTY))) {
028a63a6 1296 int i;
608514de
JK		 1297
1298 for (i = 0; i < nr_pages; i++) {
1299 struct page *page = pvec.pages[i];
1300
1301 if (unlikely(f2fs_cp_error(sbi))) {
1302 f2fs_put_page(last_page, 0);
1303 pagevec_release(&pvec);
1304 return ERR_PTR(-EIO);
1305 }
1306
1307 if (!IS_DNODE(page) || !is_cold_node(page))
1308 continue;
1309 if (ino_of_node(page) != ino)
1310 continue;
1311
1312 lock_page(page);
1313
1314 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1315continue_unlock:
1316 unlock_page(page);
1317 continue;
1318 }
1319 if (ino_of_node(page) != ino)
1320 goto continue_unlock;
1321
1322 if (!PageDirty(page)) {
1323 /* someone wrote it for us */
1324 goto continue_unlock;
1325 }
1326
1327 if (last_page)
1328 f2fs_put_page(last_page, 0);
1329
1330 get_page(page);
1331 last_page = page;
1332 unlock_page(page);
1333 }
1334 pagevec_release(&pvec);
1335 cond_resched();
1336 }
1337 return last_page;
1338}
1339
d68f735b 1340static int __write_node_page(struct page *page, bool atomic, bool *submitted,
b0af6d49
CY		 1341 struct writeback_control *wbc, bool do_balance,
1342 enum iostat_type io_type)
faa24895
JK		 1343{
1344 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1345 nid_t nid;
1346 struct node_info ni;
1347 struct f2fs_io_info fio = {
1348 .sbi = sbi,
39d787be 1349 .ino = ino_of_node(page),
faa24895
JK		 1350 .type = NODE,
1351 .op = REQ_OP_WRITE,
1352 .op_flags = wbc_to_write_flags(wbc),
1353 .page = page,
1354 .encrypted_page = NULL,
d68f735b 1355 .submitted = false,
b0af6d49 1356 .io_type = io_type,
578c6478 1357 .io_wbc = wbc,
faa24895
JK		 1358 };
1359
1360 trace_f2fs_writepage(page, NODE);
1361
db198ae0
CY		 1362 if (unlikely(f2fs_cp_error(sbi))) {
1363 dec_page_count(sbi, F2FS_DIRTY_NODES);
1364 unlock_page(page);
1365 return 0;
1366 }
1367
faa24895
JK		 1368 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1369 goto redirty_out;
faa24895
JK		 1370
1371 /* get old block addr of this node page */
1372 nid = nid_of_node(page);
1373 f2fs_bug_on(sbi, page->index != nid);
1374
1375 if (wbc->for_reclaim) {
1376 if (!down_read_trylock(&sbi->node_write))
1377 goto redirty_out;
1378 } else {
1379 down_read(&sbi->node_write);
1380 }
1381
1382 get_node_info(sbi, nid, &ni);
1383
1384 /* This page is already truncated */
1385 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1386 ClearPageUptodate(page);
1387 dec_page_count(sbi, F2FS_DIRTY_NODES);
1388 up_read(&sbi->node_write);
1389 unlock_page(page);
1390 return 0;
1391 }
1392
e7c75ab0
JK		 1393 if (atomic && !test_opt(sbi, NOBARRIER))
1394 fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
1395
faa24895 1396 set_page_writeback(page);
17c50035 1397 ClearPageError(page);
faa24895
JK		 1398 fio.old_blkaddr = ni.blk_addr;
1399 write_node_page(nid, &fio);
1400 set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
1401 dec_page_count(sbi, F2FS_DIRTY_NODES);
1402 up_read(&sbi->node_write);
1403
d68f735b 1404 if (wbc->for_reclaim) {
b9109b0e
JK		 1405 f2fs_submit_merged_write_cond(sbi, page->mapping->host, 0,
1406 page->index, NODE);
d68f735b
JK		 1407 submitted = NULL;
1408 }
faa24895
JK		 1409
1410 unlock_page(page);
1411
d68f735b 1412 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 1413 f2fs_submit_merged_write(sbi, NODE);
d68f735b
JK		 1414 submitted = NULL;
1415 }
1416 if (submitted)
1417 *submitted = fio.submitted;
faa24895 1418
401db79f
YS		 1419 if (do_balance)
1420 f2fs_balance_fs(sbi, false);
faa24895
JK		 1421 return 0;
1422
1423redirty_out:
1424 redirty_page_for_writepage(wbc, page);
1425 return AOP_WRITEPAGE_ACTIVATE;
1426}
1427
f15194fc
YH		 1428void move_node_page(struct page *node_page, int gc_type)
1429{
1430 if (gc_type == FG_GC) {
1431 struct writeback_control wbc = {
1432 .sync_mode = WB_SYNC_ALL,
1433 .nr_to_write = 1,
1434 .for_reclaim = 0,
1435 };
1436
1437 set_page_dirty(node_page);
1438 f2fs_wait_on_page_writeback(node_page, NODE, true);
1439
1440 f2fs_bug_on(F2FS_P_SB(node_page), PageWriteback(node_page));
1441 if (!clear_page_dirty_for_io(node_page))
1442 goto out_page;
1443
1444 if (__write_node_page(node_page, false, NULL,
1445 &wbc, false, FS_GC_NODE_IO))
1446 unlock_page(node_page);
1447 goto release_page;
1448 } else {
1449 /* set page dirty and write it */
1450 if (!PageWriteback(node_page))
1451 set_page_dirty(node_page);
1452 }
1453out_page:
1454 unlock_page(node_page);
1455release_page:
1456 f2fs_put_page(node_page, 0);
1457}
1458
faa24895
JK		 1459static int f2fs_write_node_page(struct page *page,
1460 struct writeback_control *wbc)
1461{
b0af6d49 1462 return __write_node_page(page, false, NULL, wbc, false, FS_NODE_IO);
faa24895
JK		 1463}
1464
26de9b11 1465int fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode,
608514de 1466 struct writeback_control *wbc, bool atomic)
e05df3b1 1467{
028a63a6 1468 pgoff_t index;
942fd319 1469 pgoff_t last_idx = ULONG_MAX;
e05df3b1 1470 struct pagevec pvec;
c267ec15 1471 int ret = 0;
608514de
JK		 1472 struct page *last_page = NULL;
1473 bool marked = false;
26de9b11 1474 nid_t ino = inode->i_ino;
028a63a6 1475 int nr_pages;
52681375 1476
608514de
JK		 1477 if (atomic) {
1478 last_page = last_fsync_dnode(sbi, ino);
1479 if (IS_ERR_OR_NULL(last_page))
1480 return PTR_ERR_OR_ZERO(last_page);
1481 }
1482retry:
86679820 1483 pagevec_init(&pvec);
52681375 1484 index = 0;
028a63a6
JK		 1485
1486 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1487 PAGECACHE_TAG_DIRTY))) {
028a63a6 1488 int i;
52681375
JK		 1489
1490 for (i = 0; i < nr_pages; i++) {
1491 struct page *page = pvec.pages[i];
d68f735b 1492 bool submitted = false;
52681375
JK		 1493
1494 if (unlikely(f2fs_cp_error(sbi))) {
608514de 1495 f2fs_put_page(last_page, 0);
52681375 1496 pagevec_release(&pvec);
9de69279
CY		 1497 ret = -EIO;
1498 goto out;
52681375
JK		 1499 }
1500
1501 if (!IS_DNODE(page) || !is_cold_node(page))
1502 continue;
1503 if (ino_of_node(page) != ino)
1504 continue;
1505
1506 lock_page(page);
1507
1508 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1509continue_unlock:
1510 unlock_page(page);
1511 continue;
1512 }
1513 if (ino_of_node(page) != ino)
1514 goto continue_unlock;
1515
608514de 1516 if (!PageDirty(page) && page != last_page) {
52681375
JK		 1517 /* someone wrote it for us */
1518 goto continue_unlock;
1519 }
1520
1521 f2fs_wait_on_page_writeback(page, NODE, true);
1522 BUG_ON(PageWriteback(page));
52681375 1523
d29fd172
JK		 1524 set_fsync_mark(page, 0);
1525 set_dentry_mark(page, 0);
1526
608514de
JK		 1527 if (!atomic || page == last_page) {
1528 set_fsync_mark(page, 1);
26de9b11
JK		 1529 if (IS_INODE(page)) {
1530 if (is_inode_flag_set(inode,
1531 FI_DIRTY_INODE))
1532 update_inode(inode, page);
608514de 1533 set_dentry_mark(page,
52681375 1534 need_dentry_mark(sbi, ino));
26de9b11 1535 }
608514de
JK		 1536 /* may be written by other thread */
1537 if (!PageDirty(page))
1538 set_page_dirty(page);
1539 }
1540
1541 if (!clear_page_dirty_for_io(page))
1542 goto continue_unlock;
52681375 1543
e7c75ab0 1544 ret = __write_node_page(page, atomic &&
d68f735b 1545 page == last_page,
b0af6d49
CY		 1546 &submitted, wbc, true,
1547 FS_NODE_IO);
c267ec15 1548 if (ret) {
52681375 1549 unlock_page(page);
608514de
JK		 1550 f2fs_put_page(last_page, 0);
1551 break;
d68f735b 1552 } else if (submitted) {
942fd319 1553 last_idx = page->index;
608514de 1554 }
3f5f4959 1555
608514de
JK		 1556 if (page == last_page) {
1557 f2fs_put_page(page, 0);
1558 marked = true;
52681375 1559 break;
c267ec15 1560 }
52681375
JK		 1561 }
1562 pagevec_release(&pvec);
1563 cond_resched();
1564
608514de 1565 if (ret || marked)
52681375
JK		 1566 break;
1567 }
608514de
JK		 1568 if (!ret && atomic && !marked) {
1569 f2fs_msg(sbi->sb, KERN_DEBUG,
1570 "Retry to write fsync mark: ino=%u, idx=%lx",
1571 ino, last_page->index);
1572 lock_page(last_page);
d40a43af 1573 f2fs_wait_on_page_writeback(last_page, NODE, true);
608514de
JK		 1574 set_page_dirty(last_page);
1575 unlock_page(last_page);
1576 goto retry;
1577 }
9de69279 1578out:
942fd319 1579 if (last_idx != ULONG_MAX)
b9109b0e 1580 f2fs_submit_merged_write_cond(sbi, NULL, ino, last_idx, NODE);
c267ec15 1581 return ret ? -EIO: 0;
52681375
JK		 1582}
1583
401db79f 1584int sync_node_pages(struct f2fs_sb_info *sbi, struct writeback_control *wbc,
b0af6d49 1585 bool do_balance, enum iostat_type io_type)
52681375 1586{
028a63a6 1587 pgoff_t index;
52681375
JK		 1588 struct pagevec pvec;
1589 int step = 0;
12bb0a8f 1590 int nwritten = 0;
3f5f4959 1591 int ret = 0;
028a63a6 1592 int nr_pages;
e05df3b1 1593
86679820 1594 pagevec_init(&pvec);
e05df3b1
JK		 1595
1596next_step:
1597 index = 0;
028a63a6
JK		 1598
1599 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1600 PAGECACHE_TAG_DIRTY))) {
028a63a6 1601 int i;
e05df3b1
JK		 1602
1603 for (i = 0; i < nr_pages; i++) {
1604 struct page *page = pvec.pages[i];
d68f735b 1605 bool submitted = false;
e05df3b1
JK		 1606
1607 /*
1608 * flushing sequence with step:
1609 * 0. indirect nodes
1610 * 1. dentry dnodes
1611 * 2. file dnodes
1612 */
1613 if (step == 0 && IS_DNODE(page))
1614 continue;
1615 if (step == 1 && (!IS_DNODE(page) ||
1616 is_cold_node(page)))
1617 continue;
1618 if (step == 2 && (!IS_DNODE(page) ||
1619 !is_cold_node(page)))
1620 continue;
9a4cbc9e 1621lock_node:
52681375 1622 if (!trylock_page(page))
e05df3b1
JK		 1623 continue;
1624
4ef51a8f 1625 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
e05df3b1
JK		 1626continue_unlock:
1627 unlock_page(page);
1628 continue;
1629 }
e05df3b1
JK		 1630
1631 if (!PageDirty(page)) {
1632 /* someone wrote it for us */
1633 goto continue_unlock;
1634 }
1635
2049d4fc 1636 /* flush inline_data */
52681375 1637 if (is_inline_node(page)) {
2049d4fc
JK		 1638 clear_inline_node(page);
1639 unlock_page(page);
1640 flush_inline_data(sbi, ino_of_node(page));
9a4cbc9e 1641 goto lock_node;
2049d4fc
JK		 1642 }
1643
fa3d2bdf
JK		 1644 f2fs_wait_on_page_writeback(page, NODE, true);
1645
1646 BUG_ON(PageWriteback(page));
e05df3b1
JK		 1647 if (!clear_page_dirty_for_io(page))
1648 goto continue_unlock;
1649
52681375
JK		 1650 set_fsync_mark(page, 0);
1651 set_dentry_mark(page, 0);
52746519 1652
401db79f 1653 ret = __write_node_page(page, false, &submitted,
b0af6d49 1654 wbc, do_balance, io_type);
d68f735b 1655 if (ret)
52746519 1656 unlock_page(page);
d68f735b 1657 else if (submitted)
3f5f4959 1658 nwritten++;
e05df3b1
JK		 1659
1660 if (--wbc->nr_to_write == 0)
1661 break;
1662 }
1663 pagevec_release(&pvec);
1664 cond_resched();
1665
1666 if (wbc->nr_to_write == 0) {
1667 step = 2;
1668 break;
1669 }
1670 }
1671
1672 if (step < 2) {
1673 step++;
1674 goto next_step;
1675 }
db198ae0 1676
3f5f4959 1677 if (nwritten)
b9109b0e 1678 f2fs_submit_merged_write(sbi, NODE);
db198ae0
CY		 1679
1680 if (unlikely(f2fs_cp_error(sbi)))
1681 return -EIO;
3f5f4959 1682 return ret;
e05df3b1
JK		 1683}
1684
cfe58f9d
JK		 1685int wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino)
1686{
028a63a6 1687 pgoff_t index = 0;
cfe58f9d 1688 struct pagevec pvec;
280db3c8 1689 int ret2, ret = 0;
028a63a6 1690 int nr_pages;
cfe58f9d 1691
86679820 1692 pagevec_init(&pvec);
4ef51a8f 1693
028a63a6 1694 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1695 PAGECACHE_TAG_WRITEBACK))) {
028a63a6 1696 int i;
cfe58f9d
JK		 1697
1698 for (i = 0; i < nr_pages; i++) {
1699 struct page *page = pvec.pages[i];
1700
4bf08ff6 1701 if (ino && ino_of_node(page) == ino) {
fec1d657 1702 f2fs_wait_on_page_writeback(page, NODE, true);
4bf08ff6
CY		 1703 if (TestClearPageError(page))
1704 ret = -EIO;
1705 }
cfe58f9d
JK		 1706 }
1707 pagevec_release(&pvec);
1708 cond_resched();
1709 }
1710
280db3c8 1711 ret2 = filemap_check_errors(NODE_MAPPING(sbi));
cfe58f9d
JK		 1712 if (!ret)
1713 ret = ret2;
1714 return ret;
1715}
1716
e05df3b1
JK		 1717static int f2fs_write_node_pages(struct address_space *mapping,
1718 struct writeback_control *wbc)
1719{
4081363f 1720 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
9dfa1baf 1721 struct blk_plug plug;
50c8cdb3 1722 long diff;
e05df3b1 1723
0771fcc7
CY		 1724 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1725 goto skip_write;
1726
4660f9c0
JK		 1727 /* balancing f2fs's metadata in background */
1728 f2fs_balance_fs_bg(sbi);
e05df3b1 1729
a7fdffbd 1730 /* collect a number of dirty node pages and write together */
87d6f890 1731 if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE))
d3baf95d 1732 goto skip_write;
a7fdffbd 1733
d31c7c3f
YH		 1734 trace_f2fs_writepages(mapping->host, wbc, NODE);
1735
50c8cdb3 1736 diff = nr_pages_to_write(sbi, NODE, wbc);
fb5566da 1737 wbc->sync_mode = WB_SYNC_NONE;
9dfa1baf 1738 blk_start_plug(&plug);
b0af6d49 1739 sync_node_pages(sbi, wbc, true, FS_NODE_IO);
9dfa1baf 1740 blk_finish_plug(&plug);
50c8cdb3 1741 wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
e05df3b1 1742 return 0;
d3baf95d
JK		 1743
1744skip_write:
1745 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES);
d31c7c3f 1746 trace_f2fs_writepages(mapping->host, wbc, NODE);
d3baf95d 1747 return 0;
e05df3b1
JK		 1748}
1749
1750static int f2fs_set_node_page_dirty(struct page *page)
1751{
26c6b887
JK		 1752 trace_f2fs_set_page_dirty(page, NODE);
1753
237c0790
JK		 1754 if (!PageUptodate(page))
1755 SetPageUptodate(page);
e05df3b1 1756 if (!PageDirty(page)) {
b87078ad 1757 __set_page_dirty_nobuffers(page);
4081363f 1758 inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
e05df3b1 1759 SetPagePrivate(page);
9e4ded3f 1760 f2fs_trace_pid(page);
e05df3b1
JK		 1761 return 1;
1762 }
1763 return 0;
1764}
1765
0a8165d7 1766/*
e05df3b1
JK		 1767 * Structure of the f2fs node operations
1768 */
1769const struct address_space_operations f2fs_node_aops = {
1770 .writepage = f2fs_write_node_page,
1771 .writepages = f2fs_write_node_pages,
1772 .set_page_dirty = f2fs_set_node_page_dirty,
487261f3
CY		 1773 .invalidatepage = f2fs_invalidate_page,
1774 .releasepage = f2fs_release_page,
5b7a487c
WG		 1775#ifdef CONFIG_MIGRATION
1776 .migratepage = f2fs_migrate_page,
1777#endif
e05df3b1
JK		 1778};
1779
8a7ed66a
JK		 1780static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i,
1781 nid_t n)
e05df3b1 1782{
8a7ed66a 1783 return radix_tree_lookup(&nm_i->free_nid_root, n);
e05df3b1
JK		 1784}
1785
9a4ffdf5 1786static int __insert_free_nid(struct f2fs_sb_info *sbi,
a0761f63 1787 struct free_nid *i, enum nid_state state)
e05df3b1 1788{
b8559dc2
CY		 1789 struct f2fs_nm_info *nm_i = NM_I(sbi);
1790
a0761f63
FL		 1791 int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i);
1792 if (err)
1793 return err;
eb0aa4b8 1794
9a4ffdf5
CY		 1795 f2fs_bug_on(sbi, state != i->state);
1796 nm_i->nid_cnt[state]++;
1797 if (state == FREE_NID)
1798 list_add_tail(&i->list, &nm_i->free_nid_list);
eb0aa4b8 1799 return 0;
b8559dc2
CY		 1800}
1801
9a4ffdf5 1802static void __remove_free_nid(struct f2fs_sb_info *sbi,
a0761f63 1803 struct free_nid *i, enum nid_state state)
b8559dc2
CY		 1804{
1805 struct f2fs_nm_info *nm_i = NM_I(sbi);
1806
9a4ffdf5
CY		 1807 f2fs_bug_on(sbi, state != i->state);
1808 nm_i->nid_cnt[state]--;
1809 if (state == FREE_NID)
1810 list_del(&i->list);
a0761f63
FL		 1811 radix_tree_delete(&nm_i->free_nid_root, i->nid);
1812}
1813
1814static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i,
1815 enum nid_state org_state, enum nid_state dst_state)
b8559dc2
CY		 1816{
1817 struct f2fs_nm_info *nm_i = NM_I(sbi);
1818
a0761f63
FL		 1819 f2fs_bug_on(sbi, org_state != i->state);
1820 i->state = dst_state;
1821 nm_i->nid_cnt[org_state]--;
1822 nm_i->nid_cnt[dst_state]++;
1823
1824 switch (dst_state) {
1825 case PREALLOC_NID:
1826 list_del(&i->list);
1827 break;
1828 case FREE_NID:
1829 list_add_tail(&i->list, &nm_i->free_nid_list);
1830 break;
1831 default:
1832 BUG_ON(1);
1833 }
e05df3b1
JK		 1834}
1835
5921aaa1
L		 1836static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid,
1837 bool set, bool build)
1838{
1839 struct f2fs_nm_info *nm_i = NM_I(sbi);
1840 unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid);
1841 unsigned int nid_ofs = nid - START_NID(nid);
1842
1843 if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap))
1844 return;
1845
1846 if (set) {
1847 if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
1848 return;
1849 __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
1850 nm_i->free_nid_count[nat_ofs]++;
1851 } else {
1852 if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
1853 return;
1854 __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
1855 if (!build)
1856 nm_i->free_nid_count[nat_ofs]--;
1857 }
1858}
1859
4ac91242 1860/* return if the nid is recognized as free */
5921aaa1
L		 1861static bool add_free_nid(struct f2fs_sb_info *sbi,
1862 nid_t nid, bool build, bool update)
e05df3b1 1863{
6fb03f3a 1864 struct f2fs_nm_info *nm_i = NM_I(sbi);
30a61ddf 1865 struct free_nid *i, *e;
59bbd474 1866 struct nat_entry *ne;
30a61ddf
CY		 1867 int err = -EINVAL;
1868 bool ret = false;
9198aceb
JK		 1869
1870 /* 0 nid should not be used */
cfb271d4 1871 if (unlikely(nid == 0))
4ac91242 1872 return false;
59bbd474 1873
7bd59381 1874 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
e05df3b1 1875 i->nid = nid;
9a4ffdf5 1876 i->state = FREE_NID;
e05df3b1 1877
5921aaa1 1878 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
769ec6e5 1879
b8559dc2 1880 spin_lock(&nm_i->nid_list_lock);
30a61ddf
CY		 1881
1882 if (build) {
1883 /*
1884 * Thread A Thread B
1885 * - f2fs_create
1886 * - f2fs_new_inode
1887 * - alloc_nid
9a4ffdf5 1888 * - __insert_nid_to_list(PREALLOC_NID)
30a61ddf
CY		 1889 * - f2fs_balance_fs_bg
1890 * - build_free_nids
1891 * - __build_free_nids
1892 * - scan_nat_page
1893 * - add_free_nid
1894 * - __lookup_nat_cache
1895 * - f2fs_add_link
1896 * - init_inode_metadata
1897 * - new_inode_page
1898 * - new_node_page
1899 * - set_node_addr
1900 * - alloc_nid_done
9a4ffdf5
CY		 1901 * - __remove_nid_from_list(PREALLOC_NID)
1902 * - __insert_nid_to_list(FREE_NID)
30a61ddf
CY		 1903 */
1904 ne = __lookup_nat_cache(nm_i, nid);
1905 if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) ||
1906 nat_get_blkaddr(ne) != NULL_ADDR))
1907 goto err_out;
1908
1909 e = __lookup_free_nid_list(nm_i, nid);
1910 if (e) {
9a4ffdf5 1911 if (e->state == FREE_NID)
30a61ddf
CY		 1912 ret = true;
1913 goto err_out;
1914 }
1915 }
1916 ret = true;
a0761f63 1917 err = __insert_free_nid(sbi, i, FREE_NID);
30a61ddf 1918err_out:
5921aaa1
L		 1919 if (update) {
1920 update_free_nid_bitmap(sbi, nid, ret, build);
1921 if (!build)
1922 nm_i->available_nids++;
1923 }
eb0aa4b8
JK		 1924 spin_unlock(&nm_i->nid_list_lock);
1925 radix_tree_preload_end();
5921aaa1 1926
30a61ddf 1927 if (err)
e05df3b1 1928 kmem_cache_free(free_nid_slab, i);
30a61ddf 1929 return ret;
e05df3b1
JK		 1930}
1931
b8559dc2 1932static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 1933{
b8559dc2 1934 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1 1935 struct free_nid *i;
cf0ee0f0
CY		 1936 bool need_free = false;
1937
b8559dc2 1938 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 1939 i = __lookup_free_nid_list(nm_i, nid);
9a4ffdf5 1940 if (i && i->state == FREE_NID) {
a0761f63 1941 __remove_free_nid(sbi, i, FREE_NID);
cf0ee0f0 1942 need_free = true;
e05df3b1 1943 }
b8559dc2 1944 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 1945
1946 if (need_free)
1947 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 1948}
1949
6fb03f3a 1950static void scan_nat_page(struct f2fs_sb_info *sbi,
e05df3b1
JK		 1951 struct page *nat_page, nid_t start_nid)
1952{
6fb03f3a 1953 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1
JK		 1954 struct f2fs_nat_block *nat_blk = page_address(nat_page);
1955 block_t blk_addr;
4ac91242 1956 unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid);
e05df3b1
JK		 1957 int i;
1958
23380b85 1959 __set_bit_le(nat_ofs, nm_i->nat_block_bitmap);
4ac91242 1960
e05df3b1
JK		 1961 i = start_nid % NAT_ENTRY_PER_BLOCK;
1962
1963 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
cfb271d4 1964 if (unlikely(start_nid >= nm_i->max_nid))
04431c44 1965 break;
23d38844
HL		 1966
1967 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
9850cf4a 1968 f2fs_bug_on(sbi, blk_addr == NEW_ADDR);
5921aaa1
L		 1969 if (blk_addr == NULL_ADDR) {
1970 add_free_nid(sbi, start_nid, true, true);
1971 } else {
1972 spin_lock(&NM_I(sbi)->nid_list_lock);
1973 update_free_nid_bitmap(sbi, start_nid, false, true);
1974 spin_unlock(&NM_I(sbi)->nid_list_lock);
1975 }
4ac91242
CY		 1976 }
1977}
1978
2fbaa25f 1979static void scan_curseg_cache(struct f2fs_sb_info *sbi)
4ac91242 1980{
4ac91242
CY		 1981 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1982 struct f2fs_journal *journal = curseg->journal;
2fbaa25f
CY		 1983 int i;
1984
1985 down_read(&curseg->journal_rwsem);
1986 for (i = 0; i < nats_in_cursum(journal); i++) {
1987 block_t addr;
1988 nid_t nid;
1989
1990 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
1991 nid = le32_to_cpu(nid_in_journal(journal, i));
1992 if (addr == NULL_ADDR)
5921aaa1 1993 add_free_nid(sbi, nid, true, false);
2fbaa25f
CY		 1994 else
1995 remove_free_nid(sbi, nid);
1996 }
1997 up_read(&curseg->journal_rwsem);
1998}
1999
2000static void scan_free_nid_bits(struct f2fs_sb_info *sbi)
2001{
2002 struct f2fs_nm_info *nm_i = NM_I(sbi);
4ac91242 2003 unsigned int i, idx;
97456574 2004 nid_t nid;
4ac91242
CY		 2005
2006 down_read(&nm_i->nat_tree_lock);
2007
2008 for (i = 0; i < nm_i->nat_blocks; i++) {
2009 if (!test_bit_le(i, nm_i->nat_block_bitmap))
2010 continue;
586d1492
CY		 2011 if (!nm_i->free_nid_count[i])
2012 continue;
4ac91242 2013 for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) {
97456574
FL		 2014 idx = find_next_bit_le(nm_i->free_nid_bitmap[i],
2015 NAT_ENTRY_PER_BLOCK, idx);
2016 if (idx >= NAT_ENTRY_PER_BLOCK)
2017 break;
4ac91242
CY		 2018
2019 nid = i * NAT_ENTRY_PER_BLOCK + idx;
5921aaa1 2020 add_free_nid(sbi, nid, true, false);
4ac91242 2021
9a4ffdf5 2022 if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS)
4ac91242
CY		 2023 goto out;
2024 }
2025 }
2026out:
2fbaa25f 2027 scan_curseg_cache(sbi);
4ac91242 2028
4ac91242 2029 up_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 2030}
2031
22ad0b6a 2032static void __build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
e05df3b1 2033{
e05df3b1 2034 struct f2fs_nm_info *nm_i = NM_I(sbi);
8760952d 2035 int i = 0;
55008d84 2036 nid_t nid = nm_i->next_scan_nid;
e05df3b1 2037
e9cdd307
YH		 2038 if (unlikely(nid >= nm_i->max_nid))
2039 nid = 0;
2040
55008d84 2041 /* Enough entries */
9a4ffdf5 2042 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
55008d84 2043 return;
e05df3b1 2044
3a2ad567 2045 if (!sync && !available_free_memory(sbi, FREE_NIDS))
55008d84 2046 return;
e05df3b1 2047
4ac91242
CY		 2048 if (!mount) {
2049 /* try to find free nids in free_nid_bitmap */
2050 scan_free_nid_bits(sbi);
2051
74986213 2052 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
4ac91242 2053 return;
22ad0b6a
JK		 2054 }
2055
55008d84 2056 /* readahead nat pages to be scanned */
26879fb1
CY		 2057 ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
2058 META_NAT, true);
e05df3b1 2059
b873b798 2060 down_read(&nm_i->nat_tree_lock);
a5131193 2061
e05df3b1 2062 while (1) {
66e83361
YH		 2063 if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
2064 nm_i->nat_block_bitmap)) {
2065 struct page *page = get_current_nat_page(sbi, nid);
e05df3b1 2066
66e83361
YH		 2067 scan_nat_page(sbi, page, nid);
2068 f2fs_put_page(page, 1);
2069 }
e05df3b1
JK		 2070
2071 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
cfb271d4 2072 if (unlikely(nid >= nm_i->max_nid))
e05df3b1 2073 nid = 0;
55008d84 2074
a6d494b6 2075 if (++i >= FREE_NID_PAGES)
e05df3b1
JK		 2076 break;
2077 }
2078
55008d84
JK		 2079 /* go to the next free nat pages to find free nids abundantly */
2080 nm_i->next_scan_nid = nid;
e05df3b1
JK		 2081
2082 /* find free nids from current sum_pages */
2fbaa25f 2083 scan_curseg_cache(sbi);
dfc08a12 2084
b873b798 2085 up_read(&nm_i->nat_tree_lock);
2db2388f
CY		 2086
2087 ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
ea1a29a0 2088 nm_i->ra_nid_pages, META_NAT, false);
e05df3b1
JK		 2089}
2090
22ad0b6a 2091void build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
2411cf5b
CY		 2092{
2093 mutex_lock(&NM_I(sbi)->build_lock);
22ad0b6a 2094 __build_free_nids(sbi, sync, mount);
2411cf5b
CY		 2095 mutex_unlock(&NM_I(sbi)->build_lock);
2096}
2097
e05df3b1
JK		 2098/*
2099 * If this function returns success, caller can obtain a new nid
2100 * from second parameter of this function.
2101 * The returned nid could be used ino as well as nid when inode is created.
2102 */
2103bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
2104{
2105 struct f2fs_nm_info *nm_i = NM_I(sbi);
2106 struct free_nid *i = NULL;
e05df3b1 2107retry:
cb78942b 2108#ifdef CONFIG_F2FS_FAULT_INJECTION
55523519
CY		 2109 if (time_to_inject(sbi, FAULT_ALLOC_NID)) {
2110 f2fs_show_injection_info(FAULT_ALLOC_NID);
cb78942b 2111 return false;
55523519 2112 }
cb78942b 2113#endif
b8559dc2 2114 spin_lock(&nm_i->nid_list_lock);
e05df3b1 2115
04d47e67
CY		 2116 if (unlikely(nm_i->available_nids == 0)) {
2117 spin_unlock(&nm_i->nid_list_lock);
2118 return false;
2119 }
e05df3b1 2120
55008d84 2121 /* We should not use stale free nids created by build_free_nids */
9a4ffdf5
CY		 2122 if (nm_i->nid_cnt[FREE_NID] && !on_build_free_nids(nm_i)) {
2123 f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list));
2124 i = list_first_entry(&nm_i->free_nid_list,
b8559dc2 2125 struct free_nid, list);
55008d84 2126 *nid = i->nid;
b8559dc2 2127
a0761f63 2128 __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID);
04d47e67 2129 nm_i->available_nids--;
4ac91242 2130
346fe752 2131 update_free_nid_bitmap(sbi, *nid, false, false);
4ac91242 2132
b8559dc2 2133 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2134 return true;
2135 }
b8559dc2 2136 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2137
2138 /* Let's scan nat pages and its caches to get free nids */
22ad0b6a 2139 build_free_nids(sbi, true, false);
55008d84 2140 goto retry;
e05df3b1
JK		 2141}
2142
0a8165d7 2143/*
e05df3b1
JK		 2144 * alloc_nid() should be called prior to this function.
2145 */
2146void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
2147{
2148 struct f2fs_nm_info *nm_i = NM_I(sbi);
2149 struct free_nid *i;
2150
b8559dc2 2151 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2152 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2 2153 f2fs_bug_on(sbi, !i);
a0761f63 2154 __remove_free_nid(sbi, i, PREALLOC_NID);
b8559dc2 2155 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2156
2157 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2158}
2159
0a8165d7 2160/*
e05df3b1
JK		 2161 * alloc_nid() should be called prior to this function.
2162 */
2163void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
2164{
49952fa1
JK		 2165 struct f2fs_nm_info *nm_i = NM_I(sbi);
2166 struct free_nid *i;
cf0ee0f0 2167 bool need_free = false;
49952fa1 2168
65985d93
JK		 2169 if (!nid)
2170 return;
2171
b8559dc2 2172 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2173 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2
CY		 2174 f2fs_bug_on(sbi, !i);
2175
6fb03f3a 2176 if (!available_free_memory(sbi, FREE_NIDS)) {
a0761f63 2177 __remove_free_nid(sbi, i, PREALLOC_NID);
cf0ee0f0 2178 need_free = true;
95630cba 2179 } else {
a0761f63 2180 __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID);
95630cba 2181 }
04d47e67
CY		 2182
2183 nm_i->available_nids++;
2184
346fe752 2185 update_free_nid_bitmap(sbi, nid, true, false);
4ac91242 2186
b8559dc2 2187 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2188
2189 if (need_free)
2190 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2191}
2192
31696580
CY		 2193int try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
2194{
2195 struct f2fs_nm_info *nm_i = NM_I(sbi);
2196 struct free_nid *i, *next;
2197 int nr = nr_shrink;
2198
9a4ffdf5 2199 if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
ad4edb83
JK		 2200 return 0;
2201
31696580
CY		 2202 if (!mutex_trylock(&nm_i->build_lock))
2203 return 0;
2204
b8559dc2 2205 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 2206 list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) {
b8559dc2 2207 if (nr_shrink <= 0 ||
9a4ffdf5 2208 nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
31696580 2209 break;
b8559dc2 2210
a0761f63 2211 __remove_free_nid(sbi, i, FREE_NID);
31696580
CY		 2212 kmem_cache_free(free_nid_slab, i);
2213 nr_shrink--;
31696580 2214 }
b8559dc2 2215 spin_unlock(&nm_i->nid_list_lock);
31696580
CY		 2216 mutex_unlock(&nm_i->build_lock);
2217
2218 return nr - nr_shrink;
2219}
2220
70cfed88 2221void recover_inline_xattr(struct inode *inode, struct page *page)
28cdce04 2222{
28cdce04
CY		 2223 void *src_addr, *dst_addr;
2224 size_t inline_size;
2225 struct page *ipage;
2226 struct f2fs_inode *ri;
2227
4081363f 2228 ipage = get_node_page(F2FS_I_SB(inode), inode->i_ino);
9850cf4a 2229 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage));
28cdce04 2230
e3b4d43f 2231 ri = F2FS_INODE(page);
1eca05aa
YH		 2232 if (ri->i_inline & F2FS_INLINE_XATTR) {
2233 set_inode_flag(inode, FI_INLINE_XATTR);
2234 } else {
91942321 2235 clear_inode_flag(inode, FI_INLINE_XATTR);
e3b4d43f
JK		 2236 goto update_inode;
2237 }
2238
6afc662e
CY		 2239 dst_addr = inline_xattr_addr(inode, ipage);
2240 src_addr = inline_xattr_addr(inode, page);
28cdce04
CY		 2241 inline_size = inline_xattr_size(inode);
2242
fec1d657 2243 f2fs_wait_on_page_writeback(ipage, NODE, true);
28cdce04 2244 memcpy(dst_addr, src_addr, inline_size);
e3b4d43f 2245update_inode:
28cdce04
CY		 2246 update_inode(inode, ipage);
2247 f2fs_put_page(ipage, 1);
2248}
2249
e17d488b 2250int recover_xattr_data(struct inode *inode, struct page *page)
abb2366c 2251{
4081363f 2252 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
abb2366c 2253 nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
87905682
YH		 2254 nid_t new_xnid;
2255 struct dnode_of_data dn;
abb2366c 2256 struct node_info ni;
d260081c 2257 struct page *xpage;
abb2366c 2258
abb2366c
JK		 2259 if (!prev_xnid)
2260 goto recover_xnid;
2261
d260081c 2262 /* 1: invalidate the previous xattr nid */
abb2366c 2263 get_node_info(sbi, prev_xnid, &ni);
abb2366c 2264 invalidate_blocks(sbi, ni.blk_addr);
000519f2 2265 dec_valid_node_count(sbi, inode, false);
479f40c4 2266 set_node_addr(sbi, &ni, NULL_ADDR, false);
abb2366c
JK		 2267
2268recover_xnid:
d260081c 2269 /* 2: update xattr nid in inode */
87905682
YH		 2270 if (!alloc_nid(sbi, &new_xnid))
2271 return -ENOSPC;
2272
2273 set_new_dnode(&dn, inode, NULL, NULL, new_xnid);
2274 xpage = new_node_page(&dn, XATTR_NODE_OFFSET);
2275 if (IS_ERR(xpage)) {
2276 alloc_nid_failed(sbi, new_xnid);
2277 return PTR_ERR(xpage);
2278 }
2279
2280 alloc_nid_done(sbi, new_xnid);
d260081c
CY		 2281 update_inode_page(inode);
2282
2283 /* 3: update and set xattr node page dirty */
87905682 2284 memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE);
abb2366c 2285
d260081c
CY		 2286 set_page_dirty(xpage);
2287 f2fs_put_page(xpage, 1);
abb2366c 2288
d260081c 2289 return 0;
abb2366c
JK		 2290}
2291
e05df3b1
JK		 2292int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
2293{
58bfaf44 2294 struct f2fs_inode *src, *dst;
e05df3b1
JK		 2295 nid_t ino = ino_of_node(page);
2296 struct node_info old_ni, new_ni;
2297 struct page *ipage;
2298
e8271fa3
JK		 2299 get_node_info(sbi, ino, &old_ni);
2300
2301 if (unlikely(old_ni.blk_addr != NULL_ADDR))
2302 return -EINVAL;
e8ea9b3d 2303retry:
300e129c 2304 ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false);
e8ea9b3d
JK		 2305 if (!ipage) {
2306 congestion_wait(BLK_RW_ASYNC, HZ/50);
2307 goto retry;
2308 }
e05df3b1 2309
e1c42045 2310 /* Should not use this inode from free nid list */
b8559dc2 2311 remove_free_nid(sbi, ino);
e05df3b1 2312
237c0790
JK		 2313 if (!PageUptodate(ipage))
2314 SetPageUptodate(ipage);
e05df3b1 2315 fill_node_footer(ipage, ino, ino, 0, true);
c5667575 2316 set_cold_node(page, false);
e05df3b1 2317
58bfaf44
JK		 2318 src = F2FS_INODE(page);
2319 dst = F2FS_INODE(ipage);
e05df3b1 2320
58bfaf44
JK		 2321 memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src);
2322 dst->i_size = 0;
2323 dst->i_blocks = cpu_to_le64(1);
2324 dst->i_links = cpu_to_le32(1);
2325 dst->i_xattr_nid = 0;
7a2af766 2326 dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR);
5c57132e 2327 if (dst->i_inline & F2FS_EXTRA_ATTR) {
7a2af766 2328 dst->i_extra_isize = src->i_extra_isize;
6afc662e
CY		 2329
2330 if (f2fs_sb_has_flexible_inline_xattr(sbi->sb) &&
2331 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2332 i_inline_xattr_size))
2333 dst->i_inline_xattr_size = src->i_inline_xattr_size;
2334
5c57132e
CY		 2335 if (f2fs_sb_has_project_quota(sbi->sb) &&
2336 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2337 i_projid))
2338 dst->i_projid = src->i_projid;
2339 }
e05df3b1
JK		 2340
2341 new_ni = old_ni;
2342 new_ni.ino = ino;
2343
0abd675e 2344 if (unlikely(inc_valid_node_count(sbi, NULL, true)))
65e5cd0a 2345 WARN_ON(1);
479f40c4 2346 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
e05df3b1 2347 inc_valid_inode_count(sbi);
617deb8c 2348 set_page_dirty(ipage);
e05df3b1
JK		 2349 f2fs_put_page(ipage, 1);
2350 return 0;
2351}
2352
c376fc0f 2353void restore_node_summary(struct f2fs_sb_info *sbi,
e05df3b1
JK		 2354 unsigned int segno, struct f2fs_summary_block *sum)
2355{
2356 struct f2fs_node *rn;
2357 struct f2fs_summary *sum_entry;
e05df3b1 2358 block_t addr;
9ecf4b80 2359 int i, idx, last_offset, nrpages;
e05df3b1
JK		 2360
2361 /* scan the node segment */
2362 last_offset = sbi->blocks_per_seg;
2363 addr = START_BLOCK(sbi, segno);
2364 sum_entry = &sum->entries[0];
2365
9ecf4b80 2366 for (i = 0; i < last_offset; i += nrpages, addr += nrpages) {
664ba972 2367 nrpages = min(last_offset - i, BIO_MAX_PAGES);
393ff91f 2368
e1c42045 2369 /* readahead node pages */
26879fb1 2370 ra_meta_pages(sbi, addr, nrpages, META_POR, true);
e05df3b1 2371
9ecf4b80 2372 for (idx = addr; idx < addr + nrpages; idx++) {
2b947003 2373 struct page *page = get_tmp_page(sbi, idx);
9af0ff1c 2374
9ecf4b80
CY		 2375 rn = F2FS_NODE(page);
2376 sum_entry->nid = rn->footer.nid;
2377 sum_entry->version = 0;
2378 sum_entry->ofs_in_node = 0;
2379 sum_entry++;
2380 f2fs_put_page(page, 1);
9af0ff1c 2381 }
bac4eef6 2382
9ecf4b80 2383 invalidate_mapping_pages(META_MAPPING(sbi), addr,
bac4eef6 2384 addr + nrpages);
e05df3b1 2385 }
e05df3b1
JK		 2386}
2387
aec71382 2388static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
e05df3b1
JK		 2389{
2390 struct f2fs_nm_info *nm_i = NM_I(sbi);
2391 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2392 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 2393 int i;
2394
b7ad7512 2395 down_write(&curseg->journal_rwsem);
dfc08a12 2396 for (i = 0; i < nats_in_cursum(journal); i++) {
e05df3b1
JK		 2397 struct nat_entry *ne;
2398 struct f2fs_nat_entry raw_ne;
dfc08a12 2399 nid_t nid = le32_to_cpu(nid_in_journal(journal, i));
e05df3b1 2400
dfc08a12 2401 raw_ne = nat_in_journal(journal, i);
9be32d72 2402
e05df3b1 2403 ne = __lookup_nat_cache(nm_i, nid);
e05df3b1 2404 if (!ne) {
12f9ef37
YH		 2405 ne = __alloc_nat_entry(nid, true);
2406 __init_nat_entry(nm_i, ne, &raw_ne, true);
e05df3b1 2407 }
04d47e67
CY		 2408
2409 /*
2410 * if a free nat in journal has not been used after last
2411 * checkpoint, we should remove it from available nids,
2412 * since later we will add it again.
2413 */
2414 if (!get_nat_flag(ne, IS_DIRTY) &&
2415 le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) {
2416 spin_lock(&nm_i->nid_list_lock);
2417 nm_i->available_nids--;
2418 spin_unlock(&nm_i->nid_list_lock);
2419 }
2420
e05df3b1 2421 __set_nat_cache_dirty(nm_i, ne);
e05df3b1 2422 }
dfc08a12 2423 update_nats_in_cursum(journal, -i);
b7ad7512 2424 up_write(&curseg->journal_rwsem);
e05df3b1
JK		 2425}
2426
309cc2b6
JK		 2427static void __adjust_nat_entry_set(struct nat_entry_set *nes,
2428 struct list_head *head, int max)
e05df3b1 2429{
309cc2b6 2430 struct nat_entry_set *cur;
e05df3b1 2431
309cc2b6
JK		 2432 if (nes->entry_cnt >= max)
2433 goto add_out;
e05df3b1 2434
309cc2b6
JK		 2435 list_for_each_entry(cur, head, set_list) {
2436 if (cur->entry_cnt >= nes->entry_cnt) {
2437 list_add(&nes->set_list, cur->set_list.prev);
2438 return;
2439 }
aec71382 2440 }
309cc2b6
JK		 2441add_out:
2442 list_add_tail(&nes->set_list, head);
2443}
e05df3b1 2444
9f7e4a2c 2445static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
22ad0b6a
JK		 2446 struct page *page)
2447{
2448 struct f2fs_nm_info *nm_i = NM_I(sbi);
2449 unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK;
2450 struct f2fs_nat_block *nat_blk = page_address(page);
2451 int valid = 0;
37a0ab2a 2452 int i = 0;
22ad0b6a
JK		 2453
2454 if (!enabled_nat_bits(sbi, NULL))
2455 return;
2456
37a0ab2a
FL		 2457 if (nat_index == 0) {
2458 valid = 1;
2459 i = 1;
2460 }
2461 for (; i < NAT_ENTRY_PER_BLOCK; i++) {
2462 if (nat_blk->entries[i].block_addr != NULL_ADDR)
22ad0b6a
JK		 2463 valid++;
2464 }
2465 if (valid == 0) {
23380b85
JK		 2466 __set_bit_le(nat_index, nm_i->empty_nat_bits);
2467 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK		 2468 return;
2469 }
2470
23380b85 2471 __clear_bit_le(nat_index, nm_i->empty_nat_bits);
22ad0b6a 2472 if (valid == NAT_ENTRY_PER_BLOCK)
23380b85 2473 __set_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a 2474 else
23380b85 2475 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK		 2476}
2477
309cc2b6 2478static void __flush_nat_entry_set(struct f2fs_sb_info *sbi,
22ad0b6a 2479 struct nat_entry_set *set, struct cp_control *cpc)
309cc2b6
JK		 2480{
2481 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2482 struct f2fs_journal *journal = curseg->journal;
309cc2b6
JK		 2483 nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK;
2484 bool to_journal = true;
2485 struct f2fs_nat_block *nat_blk;
2486 struct nat_entry *ne, *cur;
2487 struct page *page = NULL;
e05df3b1 2488
aec71382
CY		 2489 /*
2490 * there are two steps to flush nat entries:
2491 * #1, flush nat entries to journal in current hot data summary block.
2492 * #2, flush nat entries to nat page.
2493 */
22ad0b6a
JK		 2494 if (enabled_nat_bits(sbi, cpc) ||
2495 !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL))
309cc2b6
JK		 2496 to_journal = false;
2497
2498 if (to_journal) {
b7ad7512 2499 down_write(&curseg->journal_rwsem);
309cc2b6
JK		 2500 } else {
2501 page = get_next_nat_page(sbi, start_nid);
2502 nat_blk = page_address(page);
2503 f2fs_bug_on(sbi, !nat_blk);
2504 }
aec71382 2505
309cc2b6
JK		 2506 /* flush dirty nats in nat entry set */
2507 list_for_each_entry_safe(ne, cur, &set->entry_list, list) {
2508 struct f2fs_nat_entry *raw_ne;
2509 nid_t nid = nat_get_nid(ne);
2510 int offset;
2511
febeca6d 2512 f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR);
aec71382
CY		 2513
2514 if (to_journal) {
dfc08a12 2515 offset = lookup_journal_in_cursum(journal,
309cc2b6
JK		 2516 NAT_JOURNAL, nid, 1);
2517 f2fs_bug_on(sbi, offset < 0);
dfc08a12
CY		 2518 raw_ne = &nat_in_journal(journal, offset);
2519 nid_in_journal(journal, offset) = cpu_to_le32(nid);
aec71382 2520 } else {
309cc2b6 2521 raw_ne = &nat_blk->entries[nid - start_nid];
e05df3b1 2522 }
309cc2b6 2523 raw_nat_from_node_info(raw_ne, &ne->ni);
309cc2b6 2524 nat_reset_flag(ne);
0b28b71e 2525 __clear_nat_cache_dirty(NM_I(sbi), set, ne);
04d47e67 2526 if (nat_get_blkaddr(ne) == NULL_ADDR) {
5921aaa1 2527 add_free_nid(sbi, nid, false, true);
4ac91242
CY		 2528 } else {
2529 spin_lock(&NM_I(sbi)->nid_list_lock);
346fe752 2530 update_free_nid_bitmap(sbi, nid, false, false);
04d47e67
CY		 2531 spin_unlock(&NM_I(sbi)->nid_list_lock);
2532 }
309cc2b6 2533 }
e05df3b1 2534
22ad0b6a 2535 if (to_journal) {
b7ad7512 2536 up_write(&curseg->journal_rwsem);
22ad0b6a
JK		 2537 } else {
2538 __update_nat_bits(sbi, start_nid, page);
309cc2b6 2539 f2fs_put_page(page, 1);
22ad0b6a 2540 }
aec71382 2541
59c9081b
YH		 2542 /* Allow dirty nats by node block allocation in write_begin */
2543 if (!set->entry_cnt) {
2544 radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
2545 kmem_cache_free(nat_entry_set_slab, set);
2546 }
309cc2b6 2547}
aec71382 2548
309cc2b6
JK		 2549/*
2550 * This function is called during the checkpointing process.
2551 */
22ad0b6a 2552void flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
309cc2b6
JK		 2553{
2554 struct f2fs_nm_info *nm_i = NM_I(sbi);
2555 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2556 struct f2fs_journal *journal = curseg->journal;
7aed0d45 2557 struct nat_entry_set *setvec[SETVEC_SIZE];
309cc2b6
JK		 2558 struct nat_entry_set *set, *tmp;
2559 unsigned int found;
2560 nid_t set_idx = 0;
2561 LIST_HEAD(sets);
2562
20d047c8
CL		 2563 if (!nm_i->dirty_nat_cnt)
2564 return;
a5131193 2565
b873b798 2566 down_write(&nm_i->nat_tree_lock);
a5131193 2567
309cc2b6
JK		 2568 /*
2569 * if there are no enough space in journal to store dirty nat
2570 * entries, remove all entries from journal and merge them
2571 * into nat entry set.
2572 */
900f7362 2573 if (enabled_nat_bits(sbi, cpc) ||
22ad0b6a 2574 !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL))
309cc2b6
JK		 2575 remove_nats_in_journal(sbi);
2576
309cc2b6 2577 while ((found = __gang_lookup_nat_set(nm_i,
7aed0d45 2578 set_idx, SETVEC_SIZE, setvec))) {
309cc2b6
JK		 2579 unsigned idx;
2580 set_idx = setvec[found - 1]->set + 1;
2581 for (idx = 0; idx < found; idx++)
2582 __adjust_nat_entry_set(setvec[idx], &sets,
dfc08a12 2583 MAX_NAT_JENTRIES(journal));
e05df3b1 2584 }
aec71382 2585
309cc2b6
JK		 2586 /* flush dirty nats in nat entry set */
2587 list_for_each_entry_safe(set, tmp, &sets, set_list)
22ad0b6a 2588 __flush_nat_entry_set(sbi, set, cpc);
309cc2b6 2589
b873b798 2590 up_write(&nm_i->nat_tree_lock);
59c9081b 2591 /* Allow dirty nats by node block allocation in write_begin */
e05df3b1
JK		 2592}
2593
22ad0b6a
JK		 2594static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
2595{
2596 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2597 struct f2fs_nm_info *nm_i = NM_I(sbi);
2598 unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE;
2599 unsigned int i;
2600 __u64 cp_ver = cur_cp_version(ckpt);
22ad0b6a
JK		 2601 block_t nat_bits_addr;
2602
2603 if (!enabled_nat_bits(sbi, NULL))
2604 return 0;
2605
df033caf 2606 nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
acbf054d
CY		 2607 nm_i->nat_bits = f2fs_kzalloc(sbi,
2608 nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL);
22ad0b6a
JK		 2609 if (!nm_i->nat_bits)
2610 return -ENOMEM;
2611
2612 nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg -
2613 nm_i->nat_bits_blocks;
2614 for (i = 0; i < nm_i->nat_bits_blocks; i++) {
2615 struct page *page = get_meta_page(sbi, nat_bits_addr++);
2616
2617 memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS),
2618 page_address(page), F2FS_BLKSIZE);
2619 f2fs_put_page(page, 1);
2620 }
2621
ced2c7ea 2622 cp_ver |= (cur_cp_crc(ckpt) << 32);
22ad0b6a
JK		 2623 if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) {
2624 disable_nat_bits(sbi, true);
2625 return 0;
2626 }
2627
2628 nm_i->full_nat_bits = nm_i->nat_bits + 8;
2629 nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes;
2630
2631 f2fs_msg(sbi->sb, KERN_NOTICE, "Found nat_bits in checkpoint");
2632 return 0;
2633}
2634
bd80a4b9 2635static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi)
7041d5d2
CY		 2636{
2637 struct f2fs_nm_info *nm_i = NM_I(sbi);
2638 unsigned int i = 0;
2639 nid_t nid, last_nid;
2640
2641 if (!enabled_nat_bits(sbi, NULL))
2642 return;
2643
2644 for (i = 0; i < nm_i->nat_blocks; i++) {
2645 i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i);
2646 if (i >= nm_i->nat_blocks)
2647 break;
2648
2649 __set_bit_le(i, nm_i->nat_block_bitmap);
2650
2651 nid = i * NAT_ENTRY_PER_BLOCK;
f6986ede 2652 last_nid = nid + NAT_ENTRY_PER_BLOCK;
7041d5d2 2653
346fe752 2654 spin_lock(&NM_I(sbi)->nid_list_lock);
7041d5d2 2655 for (; nid < last_nid; nid++)
346fe752
CY		 2656 update_free_nid_bitmap(sbi, nid, true, true);
2657 spin_unlock(&NM_I(sbi)->nid_list_lock);
7041d5d2
CY		 2658 }
2659
2660 for (i = 0; i < nm_i->nat_blocks; i++) {
2661 i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i);
2662 if (i >= nm_i->nat_blocks)
2663 break;
2664
2665 __set_bit_le(i, nm_i->nat_block_bitmap);
2666 }
2667}
2668
e05df3b1
JK		 2669static int init_node_manager(struct f2fs_sb_info *sbi)
2670{
2671 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
2672 struct f2fs_nm_info *nm_i = NM_I(sbi);
2673 unsigned char *version_bitmap;
22ad0b6a
JK		 2674 unsigned int nat_segs;
2675 int err;
e05df3b1
JK		 2676
2677 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
2678
2679 /* segment_count_nat includes pair segment so divide to 2. */
2680 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
22ad0b6a
JK		 2681 nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
2682 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
7ee0eeab 2683
b63da15e 2684 /* not used nids: 0, node, meta, (and root counted as valid node) */
04d47e67 2685 nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count -
292c196a 2686 sbi->nquota_files - F2FS_RESERVED_NODE_NUM;
9a4ffdf5
CY		 2687 nm_i->nid_cnt[FREE_NID] = 0;
2688 nm_i->nid_cnt[PREALLOC_NID] = 0;
e05df3b1 2689 nm_i->nat_cnt = 0;
cdfc41c1 2690 nm_i->ram_thresh = DEF_RAM_THRESHOLD;
ea1a29a0 2691 nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
2304cb0c 2692 nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
e05df3b1 2693
8a7ed66a 2694 INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
9a4ffdf5 2695 INIT_LIST_HEAD(&nm_i->free_nid_list);
769ec6e5
JK		 2696 INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO);
2697 INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO);
e05df3b1 2698 INIT_LIST_HEAD(&nm_i->nat_entries);
e05df3b1
JK		 2699
2700 mutex_init(&nm_i->build_lock);
b8559dc2 2701 spin_lock_init(&nm_i->nid_list_lock);
b873b798 2702 init_rwsem(&nm_i->nat_tree_lock);
e05df3b1 2703
e05df3b1 2704 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
79b5793b 2705 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
e05df3b1
JK		 2706 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
2707 if (!version_bitmap)
2708 return -EFAULT;
2709
79b5793b
AG		 2710 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
2711 GFP_KERNEL);
2712 if (!nm_i->nat_bitmap)
2713 return -ENOMEM;
599a09b2 2714
22ad0b6a
JK		 2715 err = __get_nat_bitmaps(sbi);
2716 if (err)
2717 return err;
2718
599a09b2
CY		 2719#ifdef CONFIG_F2FS_CHECK_FS
2720 nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size,
2721 GFP_KERNEL);
2722 if (!nm_i->nat_bitmap_mir)
2723 return -ENOMEM;
2724#endif
2725
e05df3b1
JK		 2726 return 0;
2727}
2728
9f7e4a2c 2729static int init_free_nid_cache(struct f2fs_sb_info *sbi)
4ac91242
CY		 2730{
2731 struct f2fs_nm_info *nm_i = NM_I(sbi);
bb1105e4 2732 int i;
4ac91242 2733
bb1105e4
JK		 2734 nm_i->free_nid_bitmap = f2fs_kzalloc(sbi, nm_i->nat_blocks *
2735 sizeof(unsigned char *), GFP_KERNEL);
4ac91242
CY		 2736 if (!nm_i->free_nid_bitmap)
2737 return -ENOMEM;
2738
bb1105e4
JK		 2739 for (i = 0; i < nm_i->nat_blocks; i++) {
2740 nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi,
2741 NAT_ENTRY_BITMAP_SIZE_ALIGNED, GFP_KERNEL);
2742 if (!nm_i->free_nid_bitmap)
2743 return -ENOMEM;
2744 }
2745
628b3d14 2746 nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8,
4ac91242
CY		 2747 GFP_KERNEL);
2748 if (!nm_i->nat_block_bitmap)
2749 return -ENOMEM;
586d1492 2750
628b3d14 2751 nm_i->free_nid_count = f2fs_kvzalloc(sbi, nm_i->nat_blocks *
586d1492
CY		 2752 sizeof(unsigned short), GFP_KERNEL);
2753 if (!nm_i->free_nid_count)
2754 return -ENOMEM;
4ac91242
CY		 2755 return 0;
2756}
2757
e05df3b1
JK		 2758int build_node_manager(struct f2fs_sb_info *sbi)
2759{
2760 int err;
2761
acbf054d
CY		 2762 sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info),
2763 GFP_KERNEL);
e05df3b1
JK		 2764 if (!sbi->nm_info)
2765 return -ENOMEM;
2766
2767 err = init_node_manager(sbi);
2768 if (err)
2769 return err;
2770
4ac91242
CY		 2771 err = init_free_nid_cache(sbi);
2772 if (err)
2773 return err;
2774
7041d5d2
CY		 2775 /* load free nid status from nat_bits table */
2776 load_free_nid_bitmap(sbi);
2777
22ad0b6a 2778 build_free_nids(sbi, true, true);
e05df3b1
JK		 2779 return 0;
2780}
2781
2782void destroy_node_manager(struct f2fs_sb_info *sbi)
2783{
2784 struct f2fs_nm_info *nm_i = NM_I(sbi);
2785 struct free_nid *i, *next_i;
2786 struct nat_entry *natvec[NATVEC_SIZE];
7aed0d45 2787 struct nat_entry_set *setvec[SETVEC_SIZE];
e05df3b1
JK		 2788 nid_t nid = 0;
2789 unsigned int found;
2790
2791 if (!nm_i)
2792 return;
2793
2794 /* destroy free nid list */
b8559dc2 2795 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 2796 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
a0761f63 2797 __remove_free_nid(sbi, i, FREE_NID);
b8559dc2 2798 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0 2799 kmem_cache_free(free_nid_slab, i);
b8559dc2 2800 spin_lock(&nm_i->nid_list_lock);
e05df3b1 2801 }
9a4ffdf5
CY		 2802 f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]);
2803 f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]);
2804 f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list));
b8559dc2 2805 spin_unlock(&nm_i->nid_list_lock);
e05df3b1
JK		 2806
2807 /* destroy nat cache */
b873b798 2808 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 2809 while ((found = __gang_lookup_nat_cache(nm_i,
2810 nid, NATVEC_SIZE, natvec))) {
2811 unsigned idx;
7aed0d45 2812
b6ce391e
GZ		 2813 nid = nat_get_nid(natvec[found - 1]) + 1;
2814 for (idx = 0; idx < found; idx++)
2815 __del_from_nat_cache(nm_i, natvec[idx]);
e05df3b1 2816 }
9850cf4a 2817 f2fs_bug_on(sbi, nm_i->nat_cnt);
7aed0d45
JK		 2818
2819 /* destroy nat set cache */
2820 nid = 0;
2821 while ((found = __gang_lookup_nat_set(nm_i,
2822 nid, SETVEC_SIZE, setvec))) {
2823 unsigned idx;
2824
2825 nid = setvec[found - 1]->set + 1;
2826 for (idx = 0; idx < found; idx++) {
2827 /* entry_cnt is not zero, when cp_error was occurred */
2828 f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list));
2829 radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set);
2830 kmem_cache_free(nat_entry_set_slab, setvec[idx]);
2831 }
2832 }
b873b798 2833 up_write(&nm_i->nat_tree_lock);
e05df3b1 2834
4ac91242 2835 kvfree(nm_i->nat_block_bitmap);
bb1105e4
JK		 2836 if (nm_i->free_nid_bitmap) {
2837 int i;
2838
2839 for (i = 0; i < nm_i->nat_blocks; i++)
2840 kvfree(nm_i->free_nid_bitmap[i]);
2841 kfree(nm_i->free_nid_bitmap);
2842 }
586d1492 2843 kvfree(nm_i->free_nid_count);
4ac91242 2844
e05df3b1 2845 kfree(nm_i->nat_bitmap);
22ad0b6a 2846 kfree(nm_i->nat_bits);
599a09b2
CY		 2847#ifdef CONFIG_F2FS_CHECK_FS
2848 kfree(nm_i->nat_bitmap_mir);
2849#endif
e05df3b1
JK		 2850 sbi->nm_info = NULL;
2851 kfree(nm_i);
2852}
2853
6e6093a8 2854int __init create_node_manager_caches(void)
e05df3b1
JK		 2855{
2856 nat_entry_slab = f2fs_kmem_cache_create("nat_entry",
e8512d2e 2857 sizeof(struct nat_entry));
e05df3b1 2858 if (!nat_entry_slab)
aec71382 2859 goto fail;
e05df3b1
JK		 2860
2861 free_nid_slab = f2fs_kmem_cache_create("free_nid",
e8512d2e 2862 sizeof(struct free_nid));
aec71382 2863 if (!free_nid_slab)
ce3e6d25 2864 goto destroy_nat_entry;
aec71382
CY		 2865
2866 nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set",
2867 sizeof(struct nat_entry_set));
2868 if (!nat_entry_set_slab)
ce3e6d25 2869 goto destroy_free_nid;
e05df3b1 2870 return 0;
aec71382 2871
ce3e6d25 2872destroy_free_nid:
aec71382 2873 kmem_cache_destroy(free_nid_slab);
ce3e6d25 2874destroy_nat_entry:
aec71382
CY		 2875 kmem_cache_destroy(nat_entry_slab);
2876fail:
2877 return -ENOMEM;
e05df3b1
JK		 2878}
2879
2880void destroy_node_manager_caches(void)
2881{
aec71382 2882 kmem_cache_destroy(nat_entry_set_slab);
e05df3b1
JK		 2883 kmem_cache_destroy(free_nid_slab);
2884 kmem_cache_destroy(nat_entry_slab);
2885}
Linux 6.x block layer and io_uring tree(s)