projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: fix an oops in f2fs_is_compressed_page
[linux-block.git] / fs / f2fs / node.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
e05df3b1
JK		 3 * fs/f2fs/node.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
e05df3b1
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/mpage.h>
11#include <linux/backing-dev.h>
12#include <linux/blkdev.h>
13#include <linux/pagevec.h>
14#include <linux/swap.h>
15
16#include "f2fs.h"
17#include "node.h"
18#include "segment.h"
87905682 19#include "xattr.h"
9e4ded3f 20#include "trace.h"
51dd6249 21#include <trace/events/f2fs.h>
e05df3b1 22
4d57b86d 23#define on_f2fs_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock)
f978f5a0 24
e05df3b1
JK		 25static struct kmem_cache *nat_entry_slab;
26static struct kmem_cache *free_nid_slab;
aec71382 27static struct kmem_cache *nat_entry_set_slab;
50fa53ec 28static struct kmem_cache *fsync_node_entry_slab;
e05df3b1 29
a4f843bd
JK		 30/*
31 * Check whether the given nid is within node id range.
32 */
4d57b86d 33int f2fs_check_nid_range(struct f2fs_sb_info *sbi, nid_t nid)
a4f843bd
JK		 34{
35 if (unlikely(nid < F2FS_ROOT_INO(sbi) || nid >= NM_I(sbi)->max_nid)) {
36 set_sbi_flag(sbi, SBI_NEED_FSCK);
dcbb4c10
JP		 37 f2fs_warn(sbi, "%s: out-of-range nid=%x, run fsck to fix.",
38 __func__, nid);
10f966bb 39 return -EFSCORRUPTED;
a4f843bd
JK		 40 }
41 return 0;
42}
43
4d57b86d 44bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type)
cdfc41c1 45{
6fb03f3a 46 struct f2fs_nm_info *nm_i = NM_I(sbi);
cdfc41c1 47 struct sysinfo val;
e5e7ea3c 48 unsigned long avail_ram;
cdfc41c1 49 unsigned long mem_size = 0;
6fb03f3a 50 bool res = false;
cdfc41c1
JK		 51
52 si_meminfo(&val);
e5e7ea3c
JK		 53
54 /* only uses low memory */
55 avail_ram = val.totalram - val.totalhigh;
56
429511cd
CY		 57 /*
58 * give 25%, 25%, 50%, 50%, 50% memory for each components respectively
59 */
6fb03f3a 60 if (type == FREE_NIDS) {
9a4ffdf5 61 mem_size = (nm_i->nid_cnt[FREE_NID] *
b8559dc2 62 sizeof(struct free_nid)) >> PAGE_SHIFT;
e5e7ea3c 63 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
6fb03f3a 64 } else if (type == NAT_ENTRIES) {
e5e7ea3c 65 mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >>
09cbfeaf 66 PAGE_SHIFT;
e5e7ea3c 67 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
e589c2c4
JK		 68 if (excess_cached_nats(sbi))
69 res = false;
a1257023
JK		 70 } else if (type == DIRTY_DENTS) {
71 if (sbi->sb->s_bdi->wb.dirty_exceeded)
72 return false;
73 mem_size = get_pages(sbi, F2FS_DIRTY_DENTS);
74 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
e5e7ea3c
JK		 75 } else if (type == INO_ENTRIES) {
76 int i;
77
39d787be 78 for (i = 0; i < MAX_INO_ENTRY; i++)
8f73cbb7
KM		 79 mem_size += sbi->im[i].ino_num *
80 sizeof(struct ino_entry);
81 mem_size >>= PAGE_SHIFT;
e5e7ea3c 82 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
429511cd 83 } else if (type == EXTENT_CACHE) {
7441ccef
JK		 84 mem_size = (atomic_read(&sbi->total_ext_tree) *
85 sizeof(struct extent_tree) +
429511cd 86 atomic_read(&sbi->total_ext_node) *
09cbfeaf 87 sizeof(struct extent_node)) >> PAGE_SHIFT;
429511cd 88 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
57864ae5
JK		 89 } else if (type == INMEM_PAGES) {
90 /* it allows 20% / total_ram for inmemory pages */
91 mem_size = get_pages(sbi, F2FS_INMEM_PAGES);
92 res = mem_size < (val.totalram / 5);
1e84371f 93 } else {
1663cae4
JK		 94 if (!sbi->sb->s_bdi->wb.dirty_exceeded)
95 return true;
6fb03f3a
JK		 96 }
97 return res;
cdfc41c1
JK		 98}
99
e05df3b1
JK		 100static void clear_node_page_dirty(struct page *page)
101{
e05df3b1 102 if (PageDirty(page)) {
5ec2d99d 103 f2fs_clear_page_cache_dirty_tag(page);
e05df3b1 104 clear_page_dirty_for_io(page);
aec2f729 105 dec_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
e05df3b1
JK		 106 }
107 ClearPageUptodate(page);
108}
109
110static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
111{
80551d17 112 return f2fs_get_meta_page_nofail(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK		 113}
114
115static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
116{
117 struct page *src_page;
118 struct page *dst_page;
e05df3b1
JK		 119 pgoff_t dst_off;
120 void *src_addr;
121 void *dst_addr;
122 struct f2fs_nm_info *nm_i = NM_I(sbi);
123
80551d17 124 dst_off = next_nat_addr(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK		 125
126 /* get current nat block page with lock */
80551d17 127 src_page = get_current_nat_page(sbi, nid);
edc55aaf
JK		 128 if (IS_ERR(src_page))
129 return src_page;
4d57b86d 130 dst_page = f2fs_grab_meta_page(sbi, dst_off);
9850cf4a 131 f2fs_bug_on(sbi, PageDirty(src_page));
e05df3b1
JK		 132
133 src_addr = page_address(src_page);
134 dst_addr = page_address(dst_page);
09cbfeaf 135 memcpy(dst_addr, src_addr, PAGE_SIZE);
e05df3b1
JK		 136 set_page_dirty(dst_page);
137 f2fs_put_page(src_page, 1);
138
139 set_to_next_nat(nm_i, nid);
140
141 return dst_page;
142}
143
12f9ef37
YH		 144static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail)
145{
146 struct nat_entry *new;
147
148 if (no_fail)
2882d343 149 new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37 150 else
2882d343 151 new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37
YH		 152 if (new) {
153 nat_set_nid(new, nid);
154 nat_reset_flag(new);
155 }
156 return new;
157}
158
159static void __free_nat_entry(struct nat_entry *e)
160{
161 kmem_cache_free(nat_entry_slab, e);
162}
163
164/* must be locked by nat_tree_lock */
165static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
166 struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail)
167{
168 if (no_fail)
169 f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne);
170 else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne))
171 return NULL;
172
173 if (raw_ne)
174 node_info_from_raw_nat(&ne->ni, raw_ne);
22969158
CY		 175
176 spin_lock(&nm_i->nat_list_lock);
12f9ef37 177 list_add_tail(&ne->list, &nm_i->nat_entries);
22969158
CY		 178 spin_unlock(&nm_i->nat_list_lock);
179
12f9ef37
YH		 180 nm_i->nat_cnt++;
181 return ne;
182}
183
e05df3b1
JK		 184static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
185{
22969158
CY		 186 struct nat_entry *ne;
187
188 ne = radix_tree_lookup(&nm_i->nat_root, n);
189
190 /* for recent accessed nat entry, move it to tail of lru list */
191 if (ne && !get_nat_flag(ne, IS_DIRTY)) {
192 spin_lock(&nm_i->nat_list_lock);
193 if (!list_empty(&ne->list))
194 list_move_tail(&ne->list, &nm_i->nat_entries);
195 spin_unlock(&nm_i->nat_list_lock);
196 }
197
198 return ne;
e05df3b1
JK		 199}
200
201static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
202 nid_t start, unsigned int nr, struct nat_entry **ep)
203{
204 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
205}
206
207static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
208{
e05df3b1
JK		 209 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
210 nm_i->nat_cnt--;
12f9ef37 211 __free_nat_entry(e);
e05df3b1
JK		 212}
213
780de47c
CY		 214static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
215 struct nat_entry *ne)
309cc2b6
JK		 216{
217 nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid);
218 struct nat_entry_set *head;
219
309cc2b6
JK		 220 head = radix_tree_lookup(&nm_i->nat_set_root, set);
221 if (!head) {
80c54505 222 head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS);
309cc2b6
JK		 223
224 INIT_LIST_HEAD(&head->entry_list);
225 INIT_LIST_HEAD(&head->set_list);
226 head->set = set;
227 head->entry_cnt = 0;
9be32d72 228 f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
309cc2b6 229 }
780de47c
CY		 230 return head;
231}
232
233static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
234 struct nat_entry *ne)
235{
236 struct nat_entry_set *head;
237 bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR;
238
239 if (!new_ne)
240 head = __grab_nat_entry_set(nm_i, ne);
241
242 /*
243 * update entry_cnt in below condition:
244 * 1. update NEW_ADDR to valid block address;
245 * 2. update old block address to new one;
246 */
247 if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
248 !get_nat_flag(ne, IS_DIRTY)))
249 head->entry_cnt++;
250
251 set_nat_flag(ne, IS_PREALLOC, new_ne);
febeca6d
CY		 252
253 if (get_nat_flag(ne, IS_DIRTY))
254 goto refresh_list;
255
309cc2b6 256 nm_i->dirty_nat_cnt++;
309cc2b6 257 set_nat_flag(ne, IS_DIRTY, true);
febeca6d 258refresh_list:
22969158 259 spin_lock(&nm_i->nat_list_lock);
780de47c 260 if (new_ne)
febeca6d
CY		 261 list_del_init(&ne->list);
262 else
263 list_move_tail(&ne->list, &head->entry_list);
22969158 264 spin_unlock(&nm_i->nat_list_lock);
309cc2b6
JK		 265}
266
267static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
0b28b71e 268 struct nat_entry_set *set, struct nat_entry *ne)
309cc2b6 269{
22969158 270 spin_lock(&nm_i->nat_list_lock);
0b28b71e 271 list_move_tail(&ne->list, &nm_i->nat_entries);
22969158
CY		 272 spin_unlock(&nm_i->nat_list_lock);
273
0b28b71e
KM		 274 set_nat_flag(ne, IS_DIRTY, false);
275 set->entry_cnt--;
276 nm_i->dirty_nat_cnt--;
309cc2b6
JK		 277}
278
279static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
280 nid_t start, unsigned int nr, struct nat_entry_set **ep)
281{
282 return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep,
283 start, nr);
284}
285
50fa53ec
CY		 286bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page)
287{
288 return NODE_MAPPING(sbi) == page->mapping &&
289 IS_DNODE(page) && is_cold_node(page);
290}
291
292void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi)
293{
294 spin_lock_init(&sbi->fsync_node_lock);
295 INIT_LIST_HEAD(&sbi->fsync_node_list);
296 sbi->fsync_seg_id = 0;
297 sbi->fsync_node_num = 0;
298}
299
300static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi,
301 struct page *page)
302{
303 struct fsync_node_entry *fn;
304 unsigned long flags;
305 unsigned int seq_id;
306
307 fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, GFP_NOFS);
308
309 get_page(page);
310 fn->page = page;
311 INIT_LIST_HEAD(&fn->list);
312
313 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
314 list_add_tail(&fn->list, &sbi->fsync_node_list);
315 fn->seq_id = sbi->fsync_seg_id++;
316 seq_id = fn->seq_id;
317 sbi->fsync_node_num++;
318 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
319
320 return seq_id;
321}
322
323void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page)
324{
325 struct fsync_node_entry *fn;
326 unsigned long flags;
327
328 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
329 list_for_each_entry(fn, &sbi->fsync_node_list, list) {
330 if (fn->page == page) {
331 list_del(&fn->list);
332 sbi->fsync_node_num--;
333 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
334 kmem_cache_free(fsync_node_entry_slab, fn);
335 put_page(page);
336 return;
337 }
338 }
339 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
340 f2fs_bug_on(sbi, 1);
341}
342
343void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi)
344{
345 unsigned long flags;
346
347 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
348 sbi->fsync_seg_id = 0;
349 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
350}
351
4d57b86d 352int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK		 353{
354 struct f2fs_nm_info *nm_i = NM_I(sbi);
355 struct nat_entry *e;
2dcf51ab 356 bool need = false;
e05df3b1 357
b873b798 358 down_read(&nm_i->nat_tree_lock);
e05df3b1 359 e = __lookup_nat_cache(nm_i, nid);
2dcf51ab
JK		 360 if (e) {
361 if (!get_nat_flag(e, IS_CHECKPOINTED) &&
362 !get_nat_flag(e, HAS_FSYNCED_INODE))
363 need = true;
364 }
b873b798 365 up_read(&nm_i->nat_tree_lock);
2dcf51ab 366 return need;
e05df3b1
JK		 367}
368
4d57b86d 369bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
479f40c4
JK		 370{
371 struct f2fs_nm_info *nm_i = NM_I(sbi);
372 struct nat_entry *e;
2dcf51ab 373 bool is_cp = true;
479f40c4 374
b873b798 375 down_read(&nm_i->nat_tree_lock);
2dcf51ab
JK		 376 e = __lookup_nat_cache(nm_i, nid);
377 if (e && !get_nat_flag(e, IS_CHECKPOINTED))
378 is_cp = false;
b873b798 379 up_read(&nm_i->nat_tree_lock);
2dcf51ab 380 return is_cp;
479f40c4
JK		 381}
382
4d57b86d 383bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
b6fe5873
JK		 384{
385 struct f2fs_nm_info *nm_i = NM_I(sbi);
386 struct nat_entry *e;
88bd02c9 387 bool need_update = true;
b6fe5873 388
b873b798 389 down_read(&nm_i->nat_tree_lock);
88bd02c9
JK		 390 e = __lookup_nat_cache(nm_i, ino);
391 if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
392 (get_nat_flag(e, IS_CHECKPOINTED) ||
393 get_nat_flag(e, HAS_FSYNCED_INODE)))
394 need_update = false;
b873b798 395 up_read(&nm_i->nat_tree_lock);
88bd02c9 396 return need_update;
b6fe5873
JK		 397}
398
12f9ef37 399/* must be locked by nat_tree_lock */
1515aef0 400static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
e05df3b1
JK		 401 struct f2fs_nat_entry *ne)
402{
1515aef0 403 struct f2fs_nm_info *nm_i = NM_I(sbi);
12f9ef37 404 struct nat_entry *new, *e;
9be32d72 405
12f9ef37
YH		 406 new = __alloc_nat_entry(nid, false);
407 if (!new)
408 return;
409
410 down_write(&nm_i->nat_tree_lock);
e05df3b1 411 e = __lookup_nat_cache(nm_i, nid);
12f9ef37
YH		 412 if (!e)
413 e = __init_nat_entry(nm_i, new, ne, false);
414 else
0c0b471e
EB		 415 f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) ||
416 nat_get_blkaddr(e) !=
417 le32_to_cpu(ne->block_addr) ||
1515aef0 418 nat_get_version(e) != ne->version);
12f9ef37
YH		 419 up_write(&nm_i->nat_tree_lock);
420 if (e != new)
421 __free_nat_entry(new);
e05df3b1
JK		 422}
423
424static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
479f40c4 425 block_t new_blkaddr, bool fsync_done)
e05df3b1
JK		 426{
427 struct f2fs_nm_info *nm_i = NM_I(sbi);
428 struct nat_entry *e;
12f9ef37 429 struct nat_entry *new = __alloc_nat_entry(ni->nid, true);
9be32d72 430
b873b798 431 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 432 e = __lookup_nat_cache(nm_i, ni->nid);
433 if (!e) {
12f9ef37 434 e = __init_nat_entry(nm_i, new, NULL, true);
5c27f4ee 435 copy_node_info(&e->ni, ni);
9850cf4a 436 f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR);
e05df3b1
JK		 437 } else if (new_blkaddr == NEW_ADDR) {
438 /*
439 * when nid is reallocated,
440 * previous nat entry can be remained in nat cache.
441 * So, reinitialize it with new information.
442 */
5c27f4ee 443 copy_node_info(&e->ni, ni);
9850cf4a 444 f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR);
e05df3b1 445 }
12f9ef37
YH		 446 /* let's free early to reduce memory consumption */
447 if (e != new)
448 __free_nat_entry(new);
e05df3b1 449
e05df3b1 450 /* sanity check */
9850cf4a
JK		 451 f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr);
452 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR &&
e05df3b1 453 new_blkaddr == NULL_ADDR);
9850cf4a 454 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR &&
e05df3b1 455 new_blkaddr == NEW_ADDR);
93770ab7 456 f2fs_bug_on(sbi, __is_valid_data_blkaddr(nat_get_blkaddr(e)) &&
e05df3b1
JK		 457 new_blkaddr == NEW_ADDR);
458
e1c42045 459 /* increment version no as node is removed */
e05df3b1
JK		 460 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
461 unsigned char version = nat_get_version(e);
462 nat_set_version(e, inc_node_version(version));
463 }
464
465 /* change address */
466 nat_set_blkaddr(e, new_blkaddr);
93770ab7 467 if (!__is_valid_data_blkaddr(new_blkaddr))
88bd02c9 468 set_nat_flag(e, IS_CHECKPOINTED, false);
e05df3b1 469 __set_nat_cache_dirty(nm_i, e);
479f40c4
JK		 470
471 /* update fsync_mark if its inode nat entry is still alive */
d5b692b7
CY		 472 if (ni->nid != ni->ino)
473 e = __lookup_nat_cache(nm_i, ni->ino);
88bd02c9
JK		 474 if (e) {
475 if (fsync_done && ni->nid == ni->ino)
476 set_nat_flag(e, HAS_FSYNCED_INODE, true);
477 set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
478 }
b873b798 479 up_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 480}
481
4d57b86d 482int f2fs_try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
e05df3b1
JK		 483{
484 struct f2fs_nm_info *nm_i = NM_I(sbi);
1b38dc8e 485 int nr = nr_shrink;
e05df3b1 486
b873b798
JK		 487 if (!down_write_trylock(&nm_i->nat_tree_lock))
488 return 0;
e05df3b1 489
22969158
CY		 490 spin_lock(&nm_i->nat_list_lock);
491 while (nr_shrink) {
e05df3b1 492 struct nat_entry *ne;
22969158
CY		 493
494 if (list_empty(&nm_i->nat_entries))
495 break;
496
e05df3b1
JK		 497 ne = list_first_entry(&nm_i->nat_entries,
498 struct nat_entry, list);
22969158
CY		 499 list_del(&ne->list);
500 spin_unlock(&nm_i->nat_list_lock);
501
e05df3b1
JK		 502 __del_from_nat_cache(nm_i, ne);
503 nr_shrink--;
22969158
CY		 504
505 spin_lock(&nm_i->nat_list_lock);
e05df3b1 506 }
22969158
CY		 507 spin_unlock(&nm_i->nat_list_lock);
508
b873b798 509 up_write(&nm_i->nat_tree_lock);
1b38dc8e 510 return nr - nr_shrink;
e05df3b1
JK		 511}
512
7735730d 513int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
4d57b86d 514 struct node_info *ni)
e05df3b1
JK		 515{
516 struct f2fs_nm_info *nm_i = NM_I(sbi);
517 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 518 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 519 nid_t start_nid = START_NID(nid);
520 struct f2fs_nat_block *nat_blk;
521 struct page *page = NULL;
522 struct f2fs_nat_entry ne;
523 struct nat_entry *e;
66a82d1f 524 pgoff_t index;
93770ab7 525 block_t blkaddr;
e05df3b1
JK		 526 int i;
527
528 ni->nid = nid;
529
530 /* Check nat cache */
b873b798 531 down_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 532 e = __lookup_nat_cache(nm_i, nid);
533 if (e) {
534 ni->ino = nat_get_ino(e);
535 ni->blk_addr = nat_get_blkaddr(e);
536 ni->version = nat_get_version(e);
b873b798 537 up_read(&nm_i->nat_tree_lock);
7735730d 538 return 0;
1515aef0 539 }
e05df3b1 540
3547ea96
JK		 541 memset(&ne, 0, sizeof(struct f2fs_nat_entry));
542
e05df3b1 543 /* Check current segment summary */
b7ad7512 544 down_read(&curseg->journal_rwsem);
4d57b86d 545 i = f2fs_lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0);
e05df3b1 546 if (i >= 0) {
dfc08a12 547 ne = nat_in_journal(journal, i);
e05df3b1
JK		 548 node_info_from_raw_nat(ni, &ne);
549 }
b7ad7512 550 up_read(&curseg->journal_rwsem);
66a82d1f
YH		 551 if (i >= 0) {
552 up_read(&nm_i->nat_tree_lock);
e05df3b1 553 goto cache;
66a82d1f 554 }
e05df3b1
JK		 555
556 /* Fill node_info from nat page */
66a82d1f
YH		 557 index = current_nat_addr(sbi, nid);
558 up_read(&nm_i->nat_tree_lock);
559
4d57b86d 560 page = f2fs_get_meta_page(sbi, index);
7735730d
CY		 561 if (IS_ERR(page))
562 return PTR_ERR(page);
563
e05df3b1
JK		 564 nat_blk = (struct f2fs_nat_block *)page_address(page);
565 ne = nat_blk->entries[nid - start_nid];
566 node_info_from_raw_nat(ni, &ne);
567 f2fs_put_page(page, 1);
568cache:
93770ab7
CY		 569 blkaddr = le32_to_cpu(ne.block_addr);
570 if (__is_valid_data_blkaddr(blkaddr) &&
571 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE))
572 return -EFAULT;
573
e05df3b1 574 /* cache nat entry */
1515aef0 575 cache_nat_entry(sbi, nid, &ne);
7735730d 576 return 0;
e05df3b1
JK		 577}
578
79344efb
JK		 579/*
580 * readahead MAX_RA_NODE number of node pages.
581 */
4d57b86d 582static void f2fs_ra_node_pages(struct page *parent, int start, int n)
79344efb
JK		 583{
584 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
585 struct blk_plug plug;
586 int i, end;
587 nid_t nid;
588
589 blk_start_plug(&plug);
590
591 /* Then, try readahead for siblings of the desired node */
592 end = start + n;
593 end = min(end, NIDS_PER_BLOCK);
594 for (i = start; i < end; i++) {
595 nid = get_nid(parent, i, false);
4d57b86d 596 f2fs_ra_node_page(sbi, nid);
79344efb
JK		 597 }
598
599 blk_finish_plug(&plug);
600}
601
4d57b86d 602pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs)
3cf45747
CY		 603{
604 const long direct_index = ADDRS_PER_INODE(dn->inode);
d02a6e61
CY		 605 const long direct_blks = ADDRS_PER_BLOCK(dn->inode);
606 const long indirect_blks = ADDRS_PER_BLOCK(dn->inode) * NIDS_PER_BLOCK;
607 unsigned int skipped_unit = ADDRS_PER_BLOCK(dn->inode);
3cf45747
CY		 608 int cur_level = dn->cur_level;
609 int max_level = dn->max_level;
610 pgoff_t base = 0;
611
612 if (!dn->max_level)
613 return pgofs + 1;
614
615 while (max_level-- > cur_level)
616 skipped_unit *= NIDS_PER_BLOCK;
617
618 switch (dn->max_level) {
619 case 3:
620 base += 2 * indirect_blks;
0a4c9265 621 /* fall through */
3cf45747
CY		 622 case 2:
623 base += 2 * direct_blks;
0a4c9265 624 /* fall through */
3cf45747
CY		 625 case 1:
626 base += direct_index;
627 break;
628 default:
629 f2fs_bug_on(F2FS_I_SB(dn->inode), 1);
630 }
631
632 return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base;
633}
634
0a8165d7 635/*
e05df3b1
JK		 636 * The maximum depth is four.
637 * Offset[0] will have raw inode offset.
638 */
81ca7350 639static int get_node_path(struct inode *inode, long block,
de93653f 640 int offset[4], unsigned int noffset[4])
e05df3b1 641{
81ca7350 642 const long direct_index = ADDRS_PER_INODE(inode);
d02a6e61 643 const long direct_blks = ADDRS_PER_BLOCK(inode);
e05df3b1 644 const long dptrs_per_blk = NIDS_PER_BLOCK;
d02a6e61 645 const long indirect_blks = ADDRS_PER_BLOCK(inode) * NIDS_PER_BLOCK;
e05df3b1
JK		 646 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
647 int n = 0;
648 int level = 0;
649
650 noffset[0] = 0;
651
652 if (block < direct_index) {
25c0a6e5 653 offset[n] = block;
e05df3b1
JK		 654 goto got;
655 }
656 block -= direct_index;
657 if (block < direct_blks) {
658 offset[n++] = NODE_DIR1_BLOCK;
659 noffset[n] = 1;
25c0a6e5 660 offset[n] = block;
e05df3b1
JK		 661 level = 1;
662 goto got;
663 }
664 block -= direct_blks;
665 if (block < direct_blks) {
666 offset[n++] = NODE_DIR2_BLOCK;
667 noffset[n] = 2;
25c0a6e5 668 offset[n] = block;
e05df3b1
JK		 669 level = 1;
670 goto got;
671 }
672 block -= direct_blks;
673 if (block < indirect_blks) {
674 offset[n++] = NODE_IND1_BLOCK;
675 noffset[n] = 3;
676 offset[n++] = block / direct_blks;
677 noffset[n] = 4 + offset[n - 1];
25c0a6e5 678 offset[n] = block % direct_blks;
e05df3b1
JK		 679 level = 2;
680 goto got;
681 }
682 block -= indirect_blks;
683 if (block < indirect_blks) {
684 offset[n++] = NODE_IND2_BLOCK;
685 noffset[n] = 4 + dptrs_per_blk;
686 offset[n++] = block / direct_blks;
687 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
25c0a6e5 688 offset[n] = block % direct_blks;
e05df3b1
JK		 689 level = 2;
690 goto got;
691 }
692 block -= indirect_blks;
693 if (block < dindirect_blks) {
694 offset[n++] = NODE_DIND_BLOCK;
695 noffset[n] = 5 + (dptrs_per_blk * 2);
696 offset[n++] = block / indirect_blks;
697 noffset[n] = 6 + (dptrs_per_blk * 2) +
698 offset[n - 1] * (dptrs_per_blk + 1);
699 offset[n++] = (block / direct_blks) % dptrs_per_blk;
700 noffset[n] = 7 + (dptrs_per_blk * 2) +
701 offset[n - 2] * (dptrs_per_blk + 1) +
702 offset[n - 1];
25c0a6e5 703 offset[n] = block % direct_blks;
e05df3b1
JK		 704 level = 3;
705 goto got;
706 } else {
adb6dc19 707 return -E2BIG;
e05df3b1
JK		 708 }
709got:
710 return level;
711}
712
713/*
714 * Caller should call f2fs_put_dnode(dn).
4f4124d0 715 * Also, it should grab and release a rwsem by calling f2fs_lock_op() and
7a88ddb5 716 * f2fs_unlock_op() only if mode is set with ALLOC_NODE.
e05df3b1 717 */
4d57b86d 718int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
e05df3b1 719{
4081363f 720 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 721 struct page *npage[4];
f1a3b98e 722 struct page *parent = NULL;
e05df3b1
JK		 723 int offset[4];
724 unsigned int noffset[4];
725 nid_t nids[4];
3cf45747 726 int level, i = 0;
e05df3b1
JK		 727 int err = 0;
728
81ca7350 729 level = get_node_path(dn->inode, index, offset, noffset);
adb6dc19
JK		 730 if (level < 0)
731 return level;
e05df3b1
JK		 732
733 nids[0] = dn->inode->i_ino;
1646cfac 734 npage[0] = dn->inode_page;
e05df3b1 735
1646cfac 736 if (!npage[0]) {
4d57b86d 737 npage[0] = f2fs_get_node_page(sbi, nids[0]);
1646cfac
JK		 738 if (IS_ERR(npage[0]))
739 return PTR_ERR(npage[0]);
740 }
f1a3b98e
JK		 741
742 /* if inline_data is set, should not report any block indices */
743 if (f2fs_has_inline_data(dn->inode) && index) {
76629165 744 err = -ENOENT;
f1a3b98e
JK		 745 f2fs_put_page(npage[0], 1);
746 goto release_out;
747 }
748
e05df3b1 749 parent = npage[0];
52c2db3f
CL		 750 if (level != 0)
751 nids[1] = get_nid(parent, offset[0], true);
e05df3b1
JK		 752 dn->inode_page = npage[0];
753 dn->inode_page_locked = true;
754
755 /* get indirect or direct nodes */
756 for (i = 1; i <= level; i++) {
757 bool done = false;
758
266e97a8 759 if (!nids[i] && mode == ALLOC_NODE) {
e05df3b1 760 /* alloc new node */
4d57b86d 761 if (!f2fs_alloc_nid(sbi, &(nids[i]))) {
e05df3b1
JK		 762 err = -ENOSPC;
763 goto release_pages;
764 }
765
766 dn->nid = nids[i];
4d57b86d 767 npage[i] = f2fs_new_node_page(dn, noffset[i]);
e05df3b1 768 if (IS_ERR(npage[i])) {
4d57b86d 769 f2fs_alloc_nid_failed(sbi, nids[i]);
e05df3b1
JK		 770 err = PTR_ERR(npage[i]);
771 goto release_pages;
772 }
773
774 set_nid(parent, offset[i - 1], nids[i], i == 1);
4d57b86d 775 f2fs_alloc_nid_done(sbi, nids[i]);
e05df3b1 776 done = true;
266e97a8 777 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
4d57b86d 778 npage[i] = f2fs_get_node_page_ra(parent, offset[i - 1]);
e05df3b1
JK		 779 if (IS_ERR(npage[i])) {
780 err = PTR_ERR(npage[i]);
781 goto release_pages;
782 }
783 done = true;
784 }
785 if (i == 1) {
786 dn->inode_page_locked = false;
787 unlock_page(parent);
788 } else {
789 f2fs_put_page(parent, 1);
790 }
791
792 if (!done) {
4d57b86d 793 npage[i] = f2fs_get_node_page(sbi, nids[i]);
e05df3b1
JK		 794 if (IS_ERR(npage[i])) {
795 err = PTR_ERR(npage[i]);
796 f2fs_put_page(npage[0], 0);
797 goto release_out;
798 }
799 }
800 if (i < level) {
801 parent = npage[i];
802 nids[i + 1] = get_nid(parent, offset[i], false);
803 }
804 }
805 dn->nid = nids[level];
806 dn->ofs_in_node = offset[level];
807 dn->node_page = npage[level];
a2ced1ce 808 dn->data_blkaddr = f2fs_data_blkaddr(dn);
e05df3b1
JK		 809 return 0;
810
811release_pages:
812 f2fs_put_page(parent, 1);
813 if (i > 1)
814 f2fs_put_page(npage[0], 0);
815release_out:
816 dn->inode_page = NULL;
817 dn->node_page = NULL;
3cf45747
CY		 818 if (err == -ENOENT) {
819 dn->cur_level = i;
820 dn->max_level = level;
0a2aa8fb 821 dn->ofs_in_node = offset[level];
3cf45747 822 }
e05df3b1
JK		 823 return err;
824}
825
7735730d 826static int truncate_node(struct dnode_of_data *dn)
e05df3b1 827{
4081363f 828 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 829 struct node_info ni;
7735730d 830 int err;
0ea295dd 831 pgoff_t index;
e05df3b1 832
7735730d
CY		 833 err = f2fs_get_node_info(sbi, dn->nid, &ni);
834 if (err)
835 return err;
e05df3b1 836
e05df3b1 837 /* Deallocate node address */
4d57b86d 838 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 839 dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino);
479f40c4 840 set_node_addr(sbi, &ni, NULL_ADDR, false);
e05df3b1
JK		 841
842 if (dn->nid == dn->inode->i_ino) {
4d57b86d 843 f2fs_remove_orphan_inode(sbi, dn->nid);
e05df3b1 844 dec_valid_inode_count(sbi);
0f18b462 845 f2fs_inode_synced(dn->inode);
e05df3b1 846 }
000519f2 847
e05df3b1 848 clear_node_page_dirty(dn->node_page);
caf0047e 849 set_sbi_flag(sbi, SBI_IS_DIRTY);
e05df3b1 850
0ea295dd 851 index = dn->node_page->index;
e05df3b1 852 f2fs_put_page(dn->node_page, 1);
bf39c00a
JK		 853
854 invalidate_mapping_pages(NODE_MAPPING(sbi),
0ea295dd 855 index, index);
bf39c00a 856
e05df3b1 857 dn->node_page = NULL;
51dd6249 858 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
7735730d
CY		 859
860 return 0;
e05df3b1
JK		 861}
862
863static int truncate_dnode(struct dnode_of_data *dn)
864{
e05df3b1 865 struct page *page;
7735730d 866 int err;
e05df3b1
JK		 867
868 if (dn->nid == 0)
869 return 1;
870
871 /* get direct node */
4d57b86d 872 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
45586c70 873 if (PTR_ERR(page) == -ENOENT)
e05df3b1
JK		 874 return 1;
875 else if (IS_ERR(page))
876 return PTR_ERR(page);
877
878 /* Make dnode_of_data for parameter */
879 dn->node_page = page;
880 dn->ofs_in_node = 0;
4d57b86d 881 f2fs_truncate_data_blocks(dn);
7735730d
CY		 882 err = truncate_node(dn);
883 if (err)
884 return err;
885
e05df3b1
JK		 886 return 1;
887}
888
889static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
890 int ofs, int depth)
891{
e05df3b1
JK		 892 struct dnode_of_data rdn = *dn;
893 struct page *page;
894 struct f2fs_node *rn;
895 nid_t child_nid;
896 unsigned int child_nofs;
897 int freed = 0;
898 int i, ret;
899
900 if (dn->nid == 0)
901 return NIDS_PER_BLOCK + 1;
902
51dd6249
NJ		 903 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
904
4d57b86d 905 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
51dd6249
NJ		 906 if (IS_ERR(page)) {
907 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
e05df3b1 908 return PTR_ERR(page);
51dd6249 909 }
e05df3b1 910
4d57b86d 911 f2fs_ra_node_pages(page, ofs, NIDS_PER_BLOCK);
79344efb 912
45590710 913 rn = F2FS_NODE(page);
e05df3b1
JK		 914 if (depth < 3) {
915 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
916 child_nid = le32_to_cpu(rn->in.nid[i]);
917 if (child_nid == 0)
918 continue;
919 rdn.nid = child_nid;
920 ret = truncate_dnode(&rdn);
921 if (ret < 0)
922 goto out_err;
12719ae1
JK		 923 if (set_nid(page, i, 0, false))
924 dn->node_changed = true;
e05df3b1
JK		 925 }
926 } else {
927 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
928 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
929 child_nid = le32_to_cpu(rn->in.nid[i]);
930 if (child_nid == 0) {
931 child_nofs += NIDS_PER_BLOCK + 1;
932 continue;
933 }
934 rdn.nid = child_nid;
935 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
936 if (ret == (NIDS_PER_BLOCK + 1)) {
12719ae1
JK		 937 if (set_nid(page, i, 0, false))
938 dn->node_changed = true;
e05df3b1
JK		 939 child_nofs += ret;
940 } else if (ret < 0 && ret != -ENOENT) {
941 goto out_err;
942 }
943 }
944 freed = child_nofs;
945 }
946
947 if (!ofs) {
948 /* remove current indirect node */
949 dn->node_page = page;
7735730d
CY		 950 ret = truncate_node(dn);
951 if (ret)
952 goto out_err;
e05df3b1
JK		 953 freed++;
954 } else {
955 f2fs_put_page(page, 1);
956 }
51dd6249 957 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
e05df3b1
JK		 958 return freed;
959
960out_err:
961 f2fs_put_page(page, 1);
51dd6249 962 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
e05df3b1
JK		 963 return ret;
964}
965
966static int truncate_partial_nodes(struct dnode_of_data *dn,
967 struct f2fs_inode *ri, int *offset, int depth)
968{
e05df3b1
JK		 969 struct page *pages[2];
970 nid_t nid[3];
971 nid_t child_nid;
972 int err = 0;
973 int i;
974 int idx = depth - 2;
975
976 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
977 if (!nid[0])
978 return 0;
979
980 /* get indirect nodes in the path */
a225dca3 981 for (i = 0; i < idx + 1; i++) {
e1c42045 982 /* reference count'll be increased */
4d57b86d 983 pages[i] = f2fs_get_node_page(F2FS_I_SB(dn->inode), nid[i]);
e05df3b1 984 if (IS_ERR(pages[i])) {
e05df3b1 985 err = PTR_ERR(pages[i]);
a225dca3 986 idx = i - 1;
e05df3b1
JK		 987 goto fail;
988 }
989 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
990 }
991
4d57b86d 992 f2fs_ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK);
79344efb 993
e05df3b1 994 /* free direct nodes linked to a partial indirect node */
a225dca3 995 for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) {
e05df3b1
JK		 996 child_nid = get_nid(pages[idx], i, false);
997 if (!child_nid)
998 continue;
999 dn->nid = child_nid;
1000 err = truncate_dnode(dn);
1001 if (err < 0)
1002 goto fail;
12719ae1
JK		 1003 if (set_nid(pages[idx], i, 0, false))
1004 dn->node_changed = true;
e05df3b1
JK		 1005 }
1006
a225dca3 1007 if (offset[idx + 1] == 0) {
e05df3b1
JK		 1008 dn->node_page = pages[idx];
1009 dn->nid = nid[idx];
7735730d
CY		 1010 err = truncate_node(dn);
1011 if (err)
1012 goto fail;
e05df3b1
JK		 1013 } else {
1014 f2fs_put_page(pages[idx], 1);
1015 }
1016 offset[idx]++;
a225dca3 1017 offset[idx + 1] = 0;
1018 idx--;
e05df3b1 1019fail:
a225dca3 1020 for (i = idx; i >= 0; i--)
e05df3b1 1021 f2fs_put_page(pages[i], 1);
51dd6249
NJ		 1022
1023 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
1024
e05df3b1
JK		 1025 return err;
1026}
1027
0a8165d7 1028/*
e05df3b1
JK		 1029 * All the block addresses of data and nodes should be nullified.
1030 */
4d57b86d 1031int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from)
e05df3b1 1032{
4081363f 1033 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e05df3b1
JK		 1034 int err = 0, cont = 1;
1035 int level, offset[4], noffset[4];
7dd690c8 1036 unsigned int nofs = 0;
58bfaf44 1037 struct f2fs_inode *ri;
e05df3b1
JK		 1038 struct dnode_of_data dn;
1039 struct page *page;
1040
51dd6249
NJ		 1041 trace_f2fs_truncate_inode_blocks_enter(inode, from);
1042
81ca7350 1043 level = get_node_path(inode, from, offset, noffset);
adb6dc19
JK		 1044 if (level < 0)
1045 return level;
ff373558 1046
4d57b86d 1047 page = f2fs_get_node_page(sbi, inode->i_ino);
51dd6249
NJ		 1048 if (IS_ERR(page)) {
1049 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
e05df3b1 1050 return PTR_ERR(page);
51dd6249 1051 }
e05df3b1
JK		 1052
1053 set_new_dnode(&dn, inode, page, NULL, 0);
1054 unlock_page(page);
1055
58bfaf44 1056 ri = F2FS_INODE(page);
e05df3b1
JK		 1057 switch (level) {
1058 case 0:
1059 case 1:
1060 nofs = noffset[1];
1061 break;
1062 case 2:
1063 nofs = noffset[1];
1064 if (!offset[level - 1])
1065 goto skip_partial;
58bfaf44 1066 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 1067 if (err < 0 && err != -ENOENT)
1068 goto fail;
1069 nofs += 1 + NIDS_PER_BLOCK;
1070 break;
1071 case 3:
1072 nofs = 5 + 2 * NIDS_PER_BLOCK;
1073 if (!offset[level - 1])
1074 goto skip_partial;
58bfaf44 1075 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 1076 if (err < 0 && err != -ENOENT)
1077 goto fail;
1078 break;
1079 default:
1080 BUG();
1081 }
1082
1083skip_partial:
1084 while (cont) {
58bfaf44 1085 dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
e05df3b1
JK		 1086 switch (offset[0]) {
1087 case NODE_DIR1_BLOCK:
1088 case NODE_DIR2_BLOCK:
1089 err = truncate_dnode(&dn);
1090 break;
1091
1092 case NODE_IND1_BLOCK:
1093 case NODE_IND2_BLOCK:
1094 err = truncate_nodes(&dn, nofs, offset[1], 2);
1095 break;
1096
1097 case NODE_DIND_BLOCK:
1098 err = truncate_nodes(&dn, nofs, offset[1], 3);
1099 cont = 0;
1100 break;
1101
1102 default:
1103 BUG();
1104 }
1105 if (err < 0 && err != -ENOENT)
1106 goto fail;
1107 if (offset[1] == 0 &&
58bfaf44 1108 ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) {
e05df3b1 1109 lock_page(page);
ff373558 1110 BUG_ON(page->mapping != NODE_MAPPING(sbi));
bae0ee7a 1111 f2fs_wait_on_page_writeback(page, NODE, true, true);
58bfaf44 1112 ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
e05df3b1
JK		 1113 set_page_dirty(page);
1114 unlock_page(page);
1115 }
1116 offset[1] = 0;
1117 offset[0]++;
1118 nofs += err;
1119 }
1120fail:
1121 f2fs_put_page(page, 0);
51dd6249 1122 trace_f2fs_truncate_inode_blocks_exit(inode, err);
e05df3b1
JK		 1123 return err > 0 ? 0 : err;
1124}
1125
9c77f754 1126/* caller must lock inode page */
4d57b86d 1127int f2fs_truncate_xattr_node(struct inode *inode)
4f16fb0f 1128{
4081363f 1129 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4f16fb0f
JK		 1130 nid_t nid = F2FS_I(inode)->i_xattr_nid;
1131 struct dnode_of_data dn;
1132 struct page *npage;
7735730d 1133 int err;
4f16fb0f
JK		 1134
1135 if (!nid)
1136 return 0;
1137
4d57b86d 1138 npage = f2fs_get_node_page(sbi, nid);
4f16fb0f
JK		 1139 if (IS_ERR(npage))
1140 return PTR_ERR(npage);
1141
7735730d
CY		 1142 set_new_dnode(&dn, inode, NULL, npage, nid);
1143 err = truncate_node(&dn);
1144 if (err) {
1145 f2fs_put_page(npage, 1);
1146 return err;
1147 }
1148
205b9822 1149 f2fs_i_xnid_write(inode, 0);
65985d93 1150
4f16fb0f
JK		 1151 return 0;
1152}
1153
39936837 1154/*
4f4124d0
CY		 1155 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
1156 * f2fs_unlock_op().
39936837 1157 */
4d57b86d 1158int f2fs_remove_inode_page(struct inode *inode)
e05df3b1 1159{
e05df3b1 1160 struct dnode_of_data dn;
13ec7297 1161 int err;
e05df3b1 1162
c2e69583 1163 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
4d57b86d 1164 err = f2fs_get_dnode_of_data(&dn, 0, LOOKUP_NODE);
13ec7297
CY		 1165 if (err)
1166 return err;
e05df3b1 1167
4d57b86d 1168 err = f2fs_truncate_xattr_node(inode);
13ec7297 1169 if (err) {
c2e69583 1170 f2fs_put_dnode(&dn);
13ec7297 1171 return err;
e05df3b1 1172 }
c2e69583
JK		 1173
1174 /* remove potential inline_data blocks */
1175 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1176 S_ISLNK(inode->i_mode))
4d57b86d 1177 f2fs_truncate_data_blocks_range(&dn, 1);
c2e69583 1178
e1c42045 1179 /* 0 is possible, after f2fs_new_inode() has failed */
8d714f8a
JK		 1180 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
1181 f2fs_put_dnode(&dn);
1182 return -EIO;
1183 }
8b6810f8
CY		 1184
1185 if (unlikely(inode->i_blocks != 0 && inode->i_blocks != 8)) {
097a7686
CY		 1186 f2fs_warn(F2FS_I_SB(inode),
1187 "f2fs_remove_inode_page: inconsistent i_blocks, ino:%lu, iblocks:%llu",
1188 inode->i_ino, (unsigned long long)inode->i_blocks);
8b6810f8
CY		 1189 set_sbi_flag(F2FS_I_SB(inode), SBI_NEED_FSCK);
1190 }
c2e69583
JK		 1191
1192 /* will put inode & node pages */
7735730d
CY		 1193 err = truncate_node(&dn);
1194 if (err) {
1195 f2fs_put_dnode(&dn);
1196 return err;
1197 }
13ec7297 1198 return 0;
e05df3b1
JK		 1199}
1200
4d57b86d 1201struct page *f2fs_new_inode_page(struct inode *inode)
e05df3b1 1202{
e05df3b1
JK		 1203 struct dnode_of_data dn;
1204
1205 /* allocate inode page for new inode */
1206 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
44a83ff6
JK		 1207
1208 /* caller should f2fs_put_page(page, 1); */
4d57b86d 1209 return f2fs_new_node_page(&dn, 0);
e05df3b1
JK		 1210}
1211
4d57b86d 1212struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs)
e05df3b1 1213{
4081363f 1214 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
25cc5d3b 1215 struct node_info new_ni;
e05df3b1
JK		 1216 struct page *page;
1217 int err;
1218
91942321 1219 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
e05df3b1
JK		 1220 return ERR_PTR(-EPERM);
1221
300e129c 1222 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false);
e05df3b1
JK		 1223 if (!page)
1224 return ERR_PTR(-ENOMEM);
1225
0abd675e 1226 if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs))))
9c02740c 1227 goto fail;
0abd675e 1228
25cc5d3b 1229#ifdef CONFIG_F2FS_CHECK_FS
7735730d
CY		 1230 err = f2fs_get_node_info(sbi, dn->nid, &new_ni);
1231 if (err) {
1232 dec_valid_node_count(sbi, dn->inode, !ofs);
1233 goto fail;
1234 }
25cc5d3b
JK		 1235 f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR);
1236#endif
1237 new_ni.nid = dn->nid;
e05df3b1 1238 new_ni.ino = dn->inode->i_ino;
25cc5d3b
JK		 1239 new_ni.blk_addr = NULL_ADDR;
1240 new_ni.flag = 0;
1241 new_ni.version = 0;
479f40c4 1242 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
9c02740c 1243
bae0ee7a 1244 f2fs_wait_on_page_writeback(page, NODE, true, true);
9c02740c 1245 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
c5667575 1246 set_cold_node(page, S_ISDIR(dn->inode->i_mode));
237c0790
JK		 1247 if (!PageUptodate(page))
1248 SetPageUptodate(page);
12719ae1
JK		 1249 if (set_page_dirty(page))
1250 dn->node_changed = true;
e05df3b1 1251
4bc8e9bc 1252 if (f2fs_has_xattr_block(ofs))
205b9822 1253 f2fs_i_xnid_write(dn->inode, dn->nid);
479bd73a 1254
e05df3b1
JK		 1255 if (ofs == 0)
1256 inc_valid_inode_count(sbi);
e05df3b1
JK		 1257 return page;
1258
1259fail:
71e9fec5 1260 clear_node_page_dirty(page);
e05df3b1
JK		 1261 f2fs_put_page(page, 1);
1262 return ERR_PTR(err);
1263}
1264
56ae674c
JK		 1265/*
1266 * Caller should do after getting the following values.
1267 * 0: f2fs_put_page(page, 0)
86531d6b 1268 * LOCKED_PAGE or error: f2fs_put_page(page, 1)
56ae674c 1269 */
04d328de 1270static int read_node_page(struct page *page, int op_flags)
e05df3b1 1271{
4081363f 1272 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
e05df3b1 1273 struct node_info ni;
cf04e8eb 1274 struct f2fs_io_info fio = {
05ca3632 1275 .sbi = sbi,
cf04e8eb 1276 .type = NODE,
04d328de
MC		 1277 .op = REQ_OP_READ,
1278 .op_flags = op_flags,
05ca3632 1279 .page = page,
4375a336 1280 .encrypted_page = NULL,
cf04e8eb 1281 };
7735730d 1282 int err;
e05df3b1 1283
54c55c4e 1284 if (PageUptodate(page)) {
b42b179b
CY		 1285 if (!f2fs_inode_chksum_verify(sbi, page)) {
1286 ClearPageUptodate(page);
10f966bb 1287 return -EFSBADCRC;
b42b179b 1288 }
3bdad3c7 1289 return LOCKED_PAGE;
54c55c4e 1290 }
3bdad3c7 1291
7735730d
CY		 1292 err = f2fs_get_node_info(sbi, page->index, &ni);
1293 if (err)
1294 return err;
e05df3b1 1295
83a3bfdb
JK		 1296 if (unlikely(ni.blk_addr == NULL_ADDR) ||
1297 is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) {
2bca1e23 1298 ClearPageUptodate(page);
e05df3b1 1299 return -ENOENT;
393ff91f
JK		 1300 }
1301
7a9d7548 1302 fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr;
8b83ac81
CY		 1303
1304 err = f2fs_submit_page_bio(&fio);
1305
1306 if (!err)
1307 f2fs_update_iostat(sbi, FS_NODE_READ_IO, F2FS_BLKSIZE);
1308
1309 return err;
e05df3b1
JK		 1310}
1311
0a8165d7 1312/*
e05df3b1
JK		 1313 * Readahead a node page
1314 */
4d57b86d 1315void f2fs_ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 1316{
e05df3b1 1317 struct page *apage;
56ae674c 1318 int err;
e05df3b1 1319
e8458725
CY		 1320 if (!nid)
1321 return;
4d57b86d 1322 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1323 return;
e8458725 1324
5ec2d99d 1325 apage = xa_load(&NODE_MAPPING(sbi)->i_pages, nid);
999270de 1326 if (apage)
393ff91f 1327 return;
e05df3b1 1328
300e129c 1329 apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1330 if (!apage)
1331 return;
1332
70246286 1333 err = read_node_page(apage, REQ_RAHEAD);
86531d6b 1334 f2fs_put_page(apage, err ? 1 : 0);
e05df3b1
JK		 1335}
1336
17a0ee55 1337static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid,
0e022ea8 1338 struct page *parent, int start)
e05df3b1 1339{
e05df3b1 1340 struct page *page;
0e022ea8 1341 int err;
e05df3b1 1342
e05df3b1
JK		 1343 if (!nid)
1344 return ERR_PTR(-ENOENT);
4d57b86d 1345 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1346 return ERR_PTR(-EINVAL);
afcb7ca0 1347repeat:
300e129c 1348 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1349 if (!page)
1350 return ERR_PTR(-ENOMEM);
1351
70fd7614 1352 err = read_node_page(page, 0);
86531d6b
JK		 1353 if (err < 0) {
1354 f2fs_put_page(page, 1);
e05df3b1 1355 return ERR_PTR(err);
86531d6b 1356 } else if (err == LOCKED_PAGE) {
1f258ec1 1357 err = 0;
56ae674c 1358 goto page_hit;
86531d6b 1359 }
e05df3b1 1360
0e022ea8 1361 if (parent)
4d57b86d 1362 f2fs_ra_node_pages(parent, start + 1, MAX_RA_NODE);
c718379b 1363
e05df3b1 1364 lock_page(page);
0e022ea8 1365
4ef51a8f 1366 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
afcb7ca0
JK		 1367 f2fs_put_page(page, 1);
1368 goto repeat;
1369 }
1563ac75 1370
1f258ec1
CY		 1371 if (unlikely(!PageUptodate(page))) {
1372 err = -EIO;
1563ac75 1373 goto out_err;
1f258ec1 1374 }
704956ec
CY		 1375
1376 if (!f2fs_inode_chksum_verify(sbi, page)) {
10f966bb 1377 err = -EFSBADCRC;
704956ec
CY		 1378 goto out_err;
1379 }
e0f56cb4 1380page_hit:
0c9df7fb 1381 if(unlikely(nid != nid_of_node(page))) {
dcbb4c10
JP		 1382 f2fs_warn(sbi, "inconsistent node block, nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]",
1383 nid, nid_of_node(page), ino_of_node(page),
1384 ofs_of_node(page), cpver_of_node(page),
1385 next_blkaddr_of_node(page));
1f258ec1 1386 err = -EINVAL;
0c9df7fb 1387out_err:
ee605234 1388 ClearPageUptodate(page);
0c9df7fb 1389 f2fs_put_page(page, 1);
1f258ec1 1390 return ERR_PTR(err);
0c9df7fb 1391 }
e05df3b1
JK		 1392 return page;
1393}
1394
4d57b86d 1395struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
0e022ea8
CY		 1396{
1397 return __get_node_page(sbi, nid, NULL, 0);
1398}
1399
4d57b86d 1400struct page *f2fs_get_node_page_ra(struct page *parent, int start)
0e022ea8
CY		 1401{
1402 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
1403 nid_t nid = get_nid(parent, start, false);
1404
1405 return __get_node_page(sbi, nid, parent, start);
1406}
1407
2049d4fc
JK		 1408static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino)
1409{
1410 struct inode *inode;
1411 struct page *page;
0f3311a8 1412 int ret;
2049d4fc
JK		 1413
1414 /* should flush inline_data before evict_inode */
1415 inode = ilookup(sbi->sb, ino);
1416 if (!inode)
1417 return;
1418
01eccef7
CY		 1419 page = f2fs_pagecache_get_page(inode->i_mapping, 0,
1420 FGP_LOCK|FGP_NOWAIT, 0);
2049d4fc
JK		 1421 if (!page)
1422 goto iput_out;
1423
1424 if (!PageUptodate(page))
1425 goto page_out;
1426
1427 if (!PageDirty(page))
1428 goto page_out;
1429
1430 if (!clear_page_dirty_for_io(page))
1431 goto page_out;
1432
0f3311a8
CY		 1433 ret = f2fs_write_inline_data(inode, page);
1434 inode_dec_dirty_pages(inode);
4d57b86d 1435 f2fs_remove_dirty_inode(inode);
0f3311a8 1436 if (ret)
2049d4fc
JK		 1437 set_page_dirty(page);
1438page_out:
4a6de50d 1439 f2fs_put_page(page, 1);
2049d4fc
JK		 1440iput_out:
1441 iput(inode);
1442}
1443
608514de
JK		 1444static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
1445{
028a63a6 1446 pgoff_t index;
608514de
JK		 1447 struct pagevec pvec;
1448 struct page *last_page = NULL;
028a63a6 1449 int nr_pages;
608514de 1450
86679820 1451 pagevec_init(&pvec);
608514de 1452 index = 0;
028a63a6
JK		 1453
1454 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1455 PAGECACHE_TAG_DIRTY))) {
028a63a6 1456 int i;
608514de
JK		 1457
1458 for (i = 0; i < nr_pages; i++) {
1459 struct page *page = pvec.pages[i];
1460
1461 if (unlikely(f2fs_cp_error(sbi))) {
1462 f2fs_put_page(last_page, 0);
1463 pagevec_release(&pvec);
1464 return ERR_PTR(-EIO);
1465 }
1466
1467 if (!IS_DNODE(page) || !is_cold_node(page))
1468 continue;
1469 if (ino_of_node(page) != ino)
1470 continue;
1471
1472 lock_page(page);
1473
1474 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1475continue_unlock:
1476 unlock_page(page);
1477 continue;
1478 }
1479 if (ino_of_node(page) != ino)
1480 goto continue_unlock;
1481
1482 if (!PageDirty(page)) {
1483 /* someone wrote it for us */
1484 goto continue_unlock;
1485 }
1486
1487 if (last_page)
1488 f2fs_put_page(last_page, 0);
1489
1490 get_page(page);
1491 last_page = page;
1492 unlock_page(page);
1493 }
1494 pagevec_release(&pvec);
1495 cond_resched();
1496 }
1497 return last_page;
1498}
1499
d68f735b 1500static int __write_node_page(struct page *page, bool atomic, bool *submitted,
b0af6d49 1501 struct writeback_control *wbc, bool do_balance,
50fa53ec 1502 enum iostat_type io_type, unsigned int *seq_id)
faa24895
JK		 1503{
1504 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1505 nid_t nid;
1506 struct node_info ni;
1507 struct f2fs_io_info fio = {
1508 .sbi = sbi,
39d787be 1509 .ino = ino_of_node(page),
faa24895
JK		 1510 .type = NODE,
1511 .op = REQ_OP_WRITE,
1512 .op_flags = wbc_to_write_flags(wbc),
1513 .page = page,
1514 .encrypted_page = NULL,
d68f735b 1515 .submitted = false,
b0af6d49 1516 .io_type = io_type,
578c6478 1517 .io_wbc = wbc,
faa24895 1518 };
50fa53ec 1519 unsigned int seq;
faa24895
JK		 1520
1521 trace_f2fs_writepage(page, NODE);
1522
6d7c865c
JK		 1523 if (unlikely(f2fs_cp_error(sbi))) {
1524 if (is_sbi_flag_set(sbi, SBI_IS_CLOSE)) {
1525 ClearPageUptodate(page);
1526 dec_page_count(sbi, F2FS_DIRTY_NODES);
1527 unlock_page(page);
1528 return 0;
1529 }
868de613 1530 goto redirty_out;
6d7c865c 1531 }
db198ae0 1532
faa24895
JK		 1533 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1534 goto redirty_out;
faa24895 1535
100c0655
JK		 1536 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
1537 wbc->sync_mode == WB_SYNC_NONE &&
fd8c8caf
CY		 1538 IS_DNODE(page) && is_cold_node(page))
1539 goto redirty_out;
1540
faa24895
JK		 1541 /* get old block addr of this node page */
1542 nid = nid_of_node(page);
1543 f2fs_bug_on(sbi, page->index != nid);
1544
7735730d
CY		 1545 if (f2fs_get_node_info(sbi, nid, &ni))
1546 goto redirty_out;
1547
faa24895
JK		 1548 if (wbc->for_reclaim) {
1549 if (!down_read_trylock(&sbi->node_write))
1550 goto redirty_out;
1551 } else {
1552 down_read(&sbi->node_write);
1553 }
1554
faa24895
JK		 1555 /* This page is already truncated */
1556 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1557 ClearPageUptodate(page);
1558 dec_page_count(sbi, F2FS_DIRTY_NODES);
1559 up_read(&sbi->node_write);
1560 unlock_page(page);
1561 return 0;
1562 }
1563
c9b60788 1564 if (__is_valid_data_blkaddr(ni.blk_addr) &&
93770ab7
CY		 1565 !f2fs_is_valid_blkaddr(sbi, ni.blk_addr,
1566 DATA_GENERIC_ENHANCE)) {
89d13c38 1567 up_read(&sbi->node_write);
c9b60788 1568 goto redirty_out;
89d13c38 1569 }
c9b60788 1570
e7c75ab0
JK		 1571 if (atomic && !test_opt(sbi, NOBARRIER))
1572 fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
1573
dc5a9412 1574 /* should add to global list before clearing PAGECACHE status */
50fa53ec
CY		 1575 if (f2fs_in_warm_node_list(sbi, page)) {
1576 seq = f2fs_add_fsync_node_entry(sbi, page);
1577 if (seq_id)
1578 *seq_id = seq;
1579 }
1580
dc5a9412
CY		 1581 set_page_writeback(page);
1582 ClearPageError(page);
1583
faa24895 1584 fio.old_blkaddr = ni.blk_addr;
4d57b86d 1585 f2fs_do_write_node_page(nid, &fio);
faa24895
JK		 1586 set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
1587 dec_page_count(sbi, F2FS_DIRTY_NODES);
1588 up_read(&sbi->node_write);
1589
d68f735b 1590 if (wbc->for_reclaim) {
bab475c5 1591 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, NODE);
d68f735b
JK		 1592 submitted = NULL;
1593 }
faa24895
JK		 1594
1595 unlock_page(page);
1596
d68f735b 1597 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 1598 f2fs_submit_merged_write(sbi, NODE);
d68f735b
JK		 1599 submitted = NULL;
1600 }
1601 if (submitted)
1602 *submitted = fio.submitted;
faa24895 1603
401db79f
YS		 1604 if (do_balance)
1605 f2fs_balance_fs(sbi, false);
faa24895
JK		 1606 return 0;
1607
1608redirty_out:
1609 redirty_page_for_writepage(wbc, page);
1610 return AOP_WRITEPAGE_ACTIVATE;
1611}
1612
48018b4c 1613int f2fs_move_node_page(struct page *node_page, int gc_type)
f15194fc 1614{
48018b4c
CY		 1615 int err = 0;
1616
f15194fc
YH		 1617 if (gc_type == FG_GC) {
1618 struct writeback_control wbc = {
1619 .sync_mode = WB_SYNC_ALL,
1620 .nr_to_write = 1,
1621 .for_reclaim = 0,
1622 };
1623
bae0ee7a 1624 f2fs_wait_on_page_writeback(node_page, NODE, true, true);
8d64d365
CY		 1625
1626 set_page_dirty(node_page);
1627
48018b4c
CY		 1628 if (!clear_page_dirty_for_io(node_page)) {
1629 err = -EAGAIN;
f15194fc 1630 goto out_page;
48018b4c 1631 }
f15194fc
YH		 1632
1633 if (__write_node_page(node_page, false, NULL,
48018b4c
CY		 1634 &wbc, false, FS_GC_NODE_IO, NULL)) {
1635 err = -EAGAIN;
f15194fc 1636 unlock_page(node_page);
48018b4c 1637 }
f15194fc
YH		 1638 goto release_page;
1639 } else {
1640 /* set page dirty and write it */
1641 if (!PageWriteback(node_page))
1642 set_page_dirty(node_page);
1643 }
1644out_page:
1645 unlock_page(node_page);
1646release_page:
1647 f2fs_put_page(node_page, 0);
48018b4c 1648 return err;
f15194fc
YH		 1649}
1650
faa24895
JK		 1651static int f2fs_write_node_page(struct page *page,
1652 struct writeback_control *wbc)
1653{
50fa53ec
CY		 1654 return __write_node_page(page, false, NULL, wbc, false,
1655 FS_NODE_IO, NULL);
faa24895
JK		 1656}
1657
4d57b86d 1658int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode,
50fa53ec
CY		 1659 struct writeback_control *wbc, bool atomic,
1660 unsigned int *seq_id)
e05df3b1 1661{
028a63a6 1662 pgoff_t index;
e05df3b1 1663 struct pagevec pvec;
c267ec15 1664 int ret = 0;
608514de
JK		 1665 struct page *last_page = NULL;
1666 bool marked = false;
26de9b11 1667 nid_t ino = inode->i_ino;
028a63a6 1668 int nr_pages;
bab475c5 1669 int nwritten = 0;
52681375 1670
608514de
JK		 1671 if (atomic) {
1672 last_page = last_fsync_dnode(sbi, ino);
1673 if (IS_ERR_OR_NULL(last_page))
1674 return PTR_ERR_OR_ZERO(last_page);
1675 }
1676retry:
86679820 1677 pagevec_init(&pvec);
52681375 1678 index = 0;
028a63a6
JK		 1679
1680 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1681 PAGECACHE_TAG_DIRTY))) {
028a63a6 1682 int i;
52681375
JK		 1683
1684 for (i = 0; i < nr_pages; i++) {
1685 struct page *page = pvec.pages[i];
d68f735b 1686 bool submitted = false;
52681375
JK		 1687
1688 if (unlikely(f2fs_cp_error(sbi))) {
608514de 1689 f2fs_put_page(last_page, 0);
52681375 1690 pagevec_release(&pvec);
9de69279
CY		 1691 ret = -EIO;
1692 goto out;
52681375
JK		 1693 }
1694
1695 if (!IS_DNODE(page) || !is_cold_node(page))
1696 continue;
1697 if (ino_of_node(page) != ino)
1698 continue;
1699
1700 lock_page(page);
1701
1702 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1703continue_unlock:
1704 unlock_page(page);
1705 continue;
1706 }
1707 if (ino_of_node(page) != ino)
1708 goto continue_unlock;
1709
608514de 1710 if (!PageDirty(page) && page != last_page) {
52681375
JK		 1711 /* someone wrote it for us */
1712 goto continue_unlock;
1713 }
1714
bae0ee7a 1715 f2fs_wait_on_page_writeback(page, NODE, true, true);
52681375 1716
d29fd172
JK		 1717 set_fsync_mark(page, 0);
1718 set_dentry_mark(page, 0);
1719
608514de
JK		 1720 if (!atomic || page == last_page) {
1721 set_fsync_mark(page, 1);
26de9b11
JK		 1722 if (IS_INODE(page)) {
1723 if (is_inode_flag_set(inode,
1724 FI_DIRTY_INODE))
4d57b86d 1725 f2fs_update_inode(inode, page);
608514de 1726 set_dentry_mark(page,
4d57b86d 1727 f2fs_need_dentry_mark(sbi, ino));
26de9b11 1728 }
608514de
JK		 1729 /* may be written by other thread */
1730 if (!PageDirty(page))
1731 set_page_dirty(page);
1732 }
1733
1734 if (!clear_page_dirty_for_io(page))
1735 goto continue_unlock;
52681375 1736
e7c75ab0 1737 ret = __write_node_page(page, atomic &&
d68f735b 1738 page == last_page,
b0af6d49 1739 &submitted, wbc, true,
50fa53ec 1740 FS_NODE_IO, seq_id);
c267ec15 1741 if (ret) {
52681375 1742 unlock_page(page);
608514de
JK		 1743 f2fs_put_page(last_page, 0);
1744 break;
d68f735b 1745 } else if (submitted) {
bab475c5 1746 nwritten++;
608514de 1747 }
3f5f4959 1748
608514de
JK		 1749 if (page == last_page) {
1750 f2fs_put_page(page, 0);
1751 marked = true;
52681375 1752 break;
c267ec15 1753 }
52681375
JK		 1754 }
1755 pagevec_release(&pvec);
1756 cond_resched();
1757
608514de 1758 if (ret || marked)
52681375
JK		 1759 break;
1760 }
608514de 1761 if (!ret && atomic && !marked) {
dcbb4c10
JP		 1762 f2fs_debug(sbi, "Retry to write fsync mark: ino=%u, idx=%lx",
1763 ino, last_page->index);
608514de 1764 lock_page(last_page);
bae0ee7a 1765 f2fs_wait_on_page_writeback(last_page, NODE, true, true);
608514de
JK		 1766 set_page_dirty(last_page);
1767 unlock_page(last_page);
1768 goto retry;
1769 }
9de69279 1770out:
bab475c5
CY		 1771 if (nwritten)
1772 f2fs_submit_merged_write_cond(sbi, NULL, NULL, ino, NODE);
c267ec15 1773 return ret ? -EIO: 0;
52681375
JK		 1774}
1775
052a82d8
CY		 1776static int f2fs_match_ino(struct inode *inode, unsigned long ino, void *data)
1777{
1778 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1779 bool clean;
1780
1781 if (inode->i_ino != ino)
1782 return 0;
1783
1784 if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
1785 return 0;
1786
1787 spin_lock(&sbi->inode_lock[DIRTY_META]);
1788 clean = list_empty(&F2FS_I(inode)->gdirty_list);
1789 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1790
1791 if (clean)
1792 return 0;
1793
1794 inode = igrab(inode);
1795 if (!inode)
1796 return 0;
1797 return 1;
1798}
1799
1800static bool flush_dirty_inode(struct page *page)
1801{
1802 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1803 struct inode *inode;
1804 nid_t ino = ino_of_node(page);
1805
1806 inode = find_inode_nowait(sbi->sb, ino, f2fs_match_ino, NULL);
1807 if (!inode)
1808 return false;
1809
1810 f2fs_update_inode(inode, page);
1811 unlock_page(page);
1812
1813 iput(inode);
1814 return true;
1815}
1816
34c061ad
SL		 1817int f2fs_flush_inline_data(struct f2fs_sb_info *sbi)
1818{
1819 pgoff_t index = 0;
1820 struct pagevec pvec;
1821 int nr_pages;
1822 int ret = 0;
1823
1824 pagevec_init(&pvec);
1825
1826 while ((nr_pages = pagevec_lookup_tag(&pvec,
1827 NODE_MAPPING(sbi), &index, PAGECACHE_TAG_DIRTY))) {
1828 int i;
1829
1830 for (i = 0; i < nr_pages; i++) {
1831 struct page *page = pvec.pages[i];
1832
1833 if (!IS_DNODE(page))
1834 continue;
1835
1836 lock_page(page);
1837
1838 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1839continue_unlock:
1840 unlock_page(page);
1841 continue;
1842 }
1843
1844 if (!PageDirty(page)) {
1845 /* someone wrote it for us */
1846 goto continue_unlock;
1847 }
1848
1849 /* flush inline_data, if it's async context. */
1850 if (is_inline_node(page)) {
1851 clear_inline_node(page);
1852 unlock_page(page);
1853 flush_inline_data(sbi, ino_of_node(page));
1854 continue;
1855 }
1856 unlock_page(page);
1857 }
1858 pagevec_release(&pvec);
1859 cond_resched();
1860 }
1861 return ret;
1862}
1863
4d57b86d
CY		 1864int f2fs_sync_node_pages(struct f2fs_sb_info *sbi,
1865 struct writeback_control *wbc,
b0af6d49 1866 bool do_balance, enum iostat_type io_type)
52681375 1867{
028a63a6 1868 pgoff_t index;
52681375
JK		 1869 struct pagevec pvec;
1870 int step = 0;
12bb0a8f 1871 int nwritten = 0;
3f5f4959 1872 int ret = 0;
c29fd0c0 1873 int nr_pages, done = 0;
e05df3b1 1874
86679820 1875 pagevec_init(&pvec);
e05df3b1
JK		 1876
1877next_step:
1878 index = 0;
028a63a6 1879
c29fd0c0
CY		 1880 while (!done && (nr_pages = pagevec_lookup_tag(&pvec,
1881 NODE_MAPPING(sbi), &index, PAGECACHE_TAG_DIRTY))) {
028a63a6 1882 int i;
e05df3b1
JK		 1883
1884 for (i = 0; i < nr_pages; i++) {
1885 struct page *page = pvec.pages[i];
d68f735b 1886 bool submitted = false;
052a82d8 1887 bool may_dirty = true;
e05df3b1 1888
c29fd0c0
CY		 1889 /* give a priority to WB_SYNC threads */
1890 if (atomic_read(&sbi->wb_sync_req[NODE]) &&
1891 wbc->sync_mode == WB_SYNC_NONE) {
1892 done = 1;
1893 break;
1894 }
1895
e05df3b1
JK		 1896 /*
1897 * flushing sequence with step:
1898 * 0. indirect nodes
1899 * 1. dentry dnodes
1900 * 2. file dnodes
1901 */
1902 if (step == 0 && IS_DNODE(page))
1903 continue;
1904 if (step == 1 && (!IS_DNODE(page) ||
1905 is_cold_node(page)))
1906 continue;
1907 if (step == 2 && (!IS_DNODE(page) ||
1908 !is_cold_node(page)))
1909 continue;
9a4cbc9e 1910lock_node:
4b270a8c
CY		 1911 if (wbc->sync_mode == WB_SYNC_ALL)
1912 lock_page(page);
1913 else if (!trylock_page(page))
e05df3b1
JK		 1914 continue;
1915
4ef51a8f 1916 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
e05df3b1
JK		 1917continue_unlock:
1918 unlock_page(page);
1919 continue;
1920 }
e05df3b1
JK		 1921
1922 if (!PageDirty(page)) {
1923 /* someone wrote it for us */
1924 goto continue_unlock;
1925 }
1926
34c061ad
SL		 1927 /* flush inline_data, if it's async context. */
1928 if (do_balance && is_inline_node(page)) {
2049d4fc
JK		 1929 clear_inline_node(page);
1930 unlock_page(page);
1931 flush_inline_data(sbi, ino_of_node(page));
9a4cbc9e 1932 goto lock_node;
052a82d8
CY		 1933 }
1934
1935 /* flush dirty inode */
1936 if (IS_INODE(page) && may_dirty) {
1937 may_dirty = false;
1938 if (flush_dirty_inode(page))
1939 goto lock_node;
2049d4fc
JK		 1940 }
1941
bae0ee7a 1942 f2fs_wait_on_page_writeback(page, NODE, true, true);
fa3d2bdf 1943
e05df3b1
JK		 1944 if (!clear_page_dirty_for_io(page))
1945 goto continue_unlock;
1946
52681375
JK		 1947 set_fsync_mark(page, 0);
1948 set_dentry_mark(page, 0);
52746519 1949
401db79f 1950 ret = __write_node_page(page, false, &submitted,
50fa53ec 1951 wbc, do_balance, io_type, NULL);
d68f735b 1952 if (ret)
52746519 1953 unlock_page(page);
d68f735b 1954 else if (submitted)
3f5f4959 1955 nwritten++;
e05df3b1
JK		 1956
1957 if (--wbc->nr_to_write == 0)
1958 break;
1959 }
1960 pagevec_release(&pvec);
1961 cond_resched();
1962
1963 if (wbc->nr_to_write == 0) {
1964 step = 2;
1965 break;
1966 }
1967 }
1968
1969 if (step < 2) {
100c0655
JK		 1970 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
1971 wbc->sync_mode == WB_SYNC_NONE && step == 1)
fd8c8caf 1972 goto out;
e05df3b1
JK		 1973 step++;
1974 goto next_step;
1975 }
fd8c8caf 1976out:
3f5f4959 1977 if (nwritten)
b9109b0e 1978 f2fs_submit_merged_write(sbi, NODE);
db198ae0
CY		 1979
1980 if (unlikely(f2fs_cp_error(sbi)))
1981 return -EIO;
3f5f4959 1982 return ret;
e05df3b1
JK		 1983}
1984
50fa53ec
CY		 1985int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi,
1986 unsigned int seq_id)
cfe58f9d 1987{
50fa53ec
CY		 1988 struct fsync_node_entry *fn;
1989 struct page *page;
1990 struct list_head *head = &sbi->fsync_node_list;
1991 unsigned long flags;
1992 unsigned int cur_seq_id = 0;
280db3c8 1993 int ret2, ret = 0;
cfe58f9d 1994
50fa53ec
CY		 1995 while (seq_id && cur_seq_id < seq_id) {
1996 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
1997 if (list_empty(head)) {
1998 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
1999 break;
2000 }
2001 fn = list_first_entry(head, struct fsync_node_entry, list);
2002 if (fn->seq_id > seq_id) {
2003 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
2004 break;
2005 }
2006 cur_seq_id = fn->seq_id;
2007 page = fn->page;
2008 get_page(page);
2009 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
4ef51a8f 2010
bae0ee7a 2011 f2fs_wait_on_page_writeback(page, NODE, true, false);
50fa53ec
CY		 2012 if (TestClearPageError(page))
2013 ret = -EIO;
cfe58f9d 2014
50fa53ec 2015 put_page(page);
cfe58f9d 2016
50fa53ec
CY		 2017 if (ret)
2018 break;
cfe58f9d
JK		 2019 }
2020
280db3c8 2021 ret2 = filemap_check_errors(NODE_MAPPING(sbi));
cfe58f9d
JK		 2022 if (!ret)
2023 ret = ret2;
50fa53ec 2024
cfe58f9d
JK		 2025 return ret;
2026}
2027
e05df3b1
JK		 2028static int f2fs_write_node_pages(struct address_space *mapping,
2029 struct writeback_control *wbc)
2030{
4081363f 2031 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
9dfa1baf 2032 struct blk_plug plug;
50c8cdb3 2033 long diff;
e05df3b1 2034
0771fcc7
CY		 2035 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2036 goto skip_write;
2037
4660f9c0 2038 /* balancing f2fs's metadata in background */
7bcd0cfa 2039 f2fs_balance_fs_bg(sbi, true);
e05df3b1 2040
a7fdffbd 2041 /* collect a number of dirty node pages and write together */
812a9597
JK		 2042 if (wbc->sync_mode != WB_SYNC_ALL &&
2043 get_pages(sbi, F2FS_DIRTY_NODES) <
2044 nr_pages_to_skip(sbi, NODE))
d3baf95d 2045 goto skip_write;
a7fdffbd 2046
c29fd0c0
CY		 2047 if (wbc->sync_mode == WB_SYNC_ALL)
2048 atomic_inc(&sbi->wb_sync_req[NODE]);
2049 else if (atomic_read(&sbi->wb_sync_req[NODE]))
2050 goto skip_write;
2051
d31c7c3f
YH		 2052 trace_f2fs_writepages(mapping->host, wbc, NODE);
2053
50c8cdb3 2054 diff = nr_pages_to_write(sbi, NODE, wbc);
9dfa1baf 2055 blk_start_plug(&plug);
4d57b86d 2056 f2fs_sync_node_pages(sbi, wbc, true, FS_NODE_IO);
9dfa1baf 2057 blk_finish_plug(&plug);
50c8cdb3 2058 wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
c29fd0c0
CY		 2059
2060 if (wbc->sync_mode == WB_SYNC_ALL)
2061 atomic_dec(&sbi->wb_sync_req[NODE]);
e05df3b1 2062 return 0;
d3baf95d
JK		 2063
2064skip_write:
2065 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES);
d31c7c3f 2066 trace_f2fs_writepages(mapping->host, wbc, NODE);
d3baf95d 2067 return 0;
e05df3b1
JK		 2068}
2069
2070static int f2fs_set_node_page_dirty(struct page *page)
2071{
26c6b887
JK		 2072 trace_f2fs_set_page_dirty(page, NODE);
2073
237c0790
JK		 2074 if (!PageUptodate(page))
2075 SetPageUptodate(page);
54c55c4e
WG		 2076#ifdef CONFIG_F2FS_CHECK_FS
2077 if (IS_INODE(page))
2078 f2fs_inode_chksum_set(F2FS_P_SB(page), page);
2079#endif
e05df3b1 2080 if (!PageDirty(page)) {
b87078ad 2081 __set_page_dirty_nobuffers(page);
4081363f 2082 inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
240a5915 2083 f2fs_set_page_private(page, 0);
9e4ded3f 2084 f2fs_trace_pid(page);
e05df3b1
JK		 2085 return 1;
2086 }
2087 return 0;
2088}
2089
0a8165d7 2090/*
e05df3b1
JK		 2091 * Structure of the f2fs node operations
2092 */
2093const struct address_space_operations f2fs_node_aops = {
2094 .writepage = f2fs_write_node_page,
2095 .writepages = f2fs_write_node_pages,
2096 .set_page_dirty = f2fs_set_node_page_dirty,
487261f3
CY		 2097 .invalidatepage = f2fs_invalidate_page,
2098 .releasepage = f2fs_release_page,
5b7a487c
WG		 2099#ifdef CONFIG_MIGRATION
2100 .migratepage = f2fs_migrate_page,
2101#endif
e05df3b1
JK		 2102};
2103
8a7ed66a
JK		 2104static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i,
2105 nid_t n)
e05df3b1 2106{
8a7ed66a 2107 return radix_tree_lookup(&nm_i->free_nid_root, n);
e05df3b1
JK		 2108}
2109
9a4ffdf5 2110static int __insert_free_nid(struct f2fs_sb_info *sbi,
b815bdc7 2111 struct free_nid *i)
e05df3b1 2112{
b8559dc2
CY		 2113 struct f2fs_nm_info *nm_i = NM_I(sbi);
2114
a0761f63
FL		 2115 int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i);
2116 if (err)
2117 return err;
eb0aa4b8 2118
b815bdc7
LS		 2119 nm_i->nid_cnt[FREE_NID]++;
2120 list_add_tail(&i->list, &nm_i->free_nid_list);
eb0aa4b8 2121 return 0;
b8559dc2
CY		 2122}
2123
9a4ffdf5 2124static void __remove_free_nid(struct f2fs_sb_info *sbi,
a0761f63 2125 struct free_nid *i, enum nid_state state)
b8559dc2
CY		 2126{
2127 struct f2fs_nm_info *nm_i = NM_I(sbi);
2128
9a4ffdf5
CY		 2129 f2fs_bug_on(sbi, state != i->state);
2130 nm_i->nid_cnt[state]--;
2131 if (state == FREE_NID)
2132 list_del(&i->list);
a0761f63
FL		 2133 radix_tree_delete(&nm_i->free_nid_root, i->nid);
2134}
2135
2136static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i,
2137 enum nid_state org_state, enum nid_state dst_state)
b8559dc2
CY		 2138{
2139 struct f2fs_nm_info *nm_i = NM_I(sbi);
2140
a0761f63
FL		 2141 f2fs_bug_on(sbi, org_state != i->state);
2142 i->state = dst_state;
2143 nm_i->nid_cnt[org_state]--;
2144 nm_i->nid_cnt[dst_state]++;
2145
2146 switch (dst_state) {
2147 case PREALLOC_NID:
2148 list_del(&i->list);
2149 break;
2150 case FREE_NID:
2151 list_add_tail(&i->list, &nm_i->free_nid_list);
2152 break;
2153 default:
2154 BUG_ON(1);
2155 }
e05df3b1
JK		 2156}
2157
5921aaa1
L		 2158static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid,
2159 bool set, bool build)
2160{
2161 struct f2fs_nm_info *nm_i = NM_I(sbi);
2162 unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid);
2163 unsigned int nid_ofs = nid - START_NID(nid);
2164
2165 if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap))
2166 return;
2167
2168 if (set) {
2169 if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2170 return;
2171 __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2172 nm_i->free_nid_count[nat_ofs]++;
2173 } else {
2174 if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2175 return;
2176 __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2177 if (!build)
2178 nm_i->free_nid_count[nat_ofs]--;
2179 }
2180}
2181
4ac91242 2182/* return if the nid is recognized as free */
5921aaa1
L		 2183static bool add_free_nid(struct f2fs_sb_info *sbi,
2184 nid_t nid, bool build, bool update)
e05df3b1 2185{
6fb03f3a 2186 struct f2fs_nm_info *nm_i = NM_I(sbi);
30a61ddf 2187 struct free_nid *i, *e;
59bbd474 2188 struct nat_entry *ne;
30a61ddf
CY		 2189 int err = -EINVAL;
2190 bool ret = false;
9198aceb
JK		 2191
2192 /* 0 nid should not be used */
cfb271d4 2193 if (unlikely(nid == 0))
4ac91242 2194 return false;
59bbd474 2195
626bcf2b
CY		 2196 if (unlikely(f2fs_check_nid_range(sbi, nid)))
2197 return false;
2198
7bd59381 2199 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
e05df3b1 2200 i->nid = nid;
9a4ffdf5 2201 i->state = FREE_NID;
e05df3b1 2202
5921aaa1 2203 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
769ec6e5 2204
b8559dc2 2205 spin_lock(&nm_i->nid_list_lock);
30a61ddf
CY		 2206
2207 if (build) {
2208 /*
2209 * Thread A Thread B
2210 * - f2fs_create
2211 * - f2fs_new_inode
4d57b86d 2212 * - f2fs_alloc_nid
9a4ffdf5 2213 * - __insert_nid_to_list(PREALLOC_NID)
30a61ddf 2214 * - f2fs_balance_fs_bg
4d57b86d
CY		 2215 * - f2fs_build_free_nids
2216 * - __f2fs_build_free_nids
30a61ddf
CY		 2217 * - scan_nat_page
2218 * - add_free_nid
2219 * - __lookup_nat_cache
2220 * - f2fs_add_link
4d57b86d
CY		 2221 * - f2fs_init_inode_metadata
2222 * - f2fs_new_inode_page
2223 * - f2fs_new_node_page
30a61ddf 2224 * - set_node_addr
4d57b86d 2225 * - f2fs_alloc_nid_done
9a4ffdf5
CY		 2226 * - __remove_nid_from_list(PREALLOC_NID)
2227 * - __insert_nid_to_list(FREE_NID)
30a61ddf
CY		 2228 */
2229 ne = __lookup_nat_cache(nm_i, nid);
2230 if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) ||
2231 nat_get_blkaddr(ne) != NULL_ADDR))
2232 goto err_out;
2233
2234 e = __lookup_free_nid_list(nm_i, nid);
2235 if (e) {
9a4ffdf5 2236 if (e->state == FREE_NID)
30a61ddf
CY		 2237 ret = true;
2238 goto err_out;
2239 }
2240 }
2241 ret = true;
b815bdc7 2242 err = __insert_free_nid(sbi, i);
30a61ddf 2243err_out:
5921aaa1
L		 2244 if (update) {
2245 update_free_nid_bitmap(sbi, nid, ret, build);
2246 if (!build)
2247 nm_i->available_nids++;
2248 }
eb0aa4b8
JK		 2249 spin_unlock(&nm_i->nid_list_lock);
2250 radix_tree_preload_end();
5921aaa1 2251
30a61ddf 2252 if (err)
e05df3b1 2253 kmem_cache_free(free_nid_slab, i);
30a61ddf 2254 return ret;
e05df3b1
JK		 2255}
2256
b8559dc2 2257static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2258{
b8559dc2 2259 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1 2260 struct free_nid *i;
cf0ee0f0
CY		 2261 bool need_free = false;
2262
b8559dc2 2263 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2264 i = __lookup_free_nid_list(nm_i, nid);
9a4ffdf5 2265 if (i && i->state == FREE_NID) {
a0761f63 2266 __remove_free_nid(sbi, i, FREE_NID);
cf0ee0f0 2267 need_free = true;
e05df3b1 2268 }
b8559dc2 2269 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2270
2271 if (need_free)
2272 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2273}
2274
e2374015 2275static int scan_nat_page(struct f2fs_sb_info *sbi,
e05df3b1
JK		 2276 struct page *nat_page, nid_t start_nid)
2277{
6fb03f3a 2278 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1
JK		 2279 struct f2fs_nat_block *nat_blk = page_address(nat_page);
2280 block_t blk_addr;
4ac91242 2281 unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid);
e05df3b1
JK		 2282 int i;
2283
23380b85 2284 __set_bit_le(nat_ofs, nm_i->nat_block_bitmap);
4ac91242 2285
e05df3b1
JK		 2286 i = start_nid % NAT_ENTRY_PER_BLOCK;
2287
2288 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
cfb271d4 2289 if (unlikely(start_nid >= nm_i->max_nid))
04431c44 2290 break;
23d38844
HL		 2291
2292 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
e2374015
CY		 2293
2294 if (blk_addr == NEW_ADDR)
2295 return -EINVAL;
2296
5921aaa1
L		 2297 if (blk_addr == NULL_ADDR) {
2298 add_free_nid(sbi, start_nid, true, true);
2299 } else {
2300 spin_lock(&NM_I(sbi)->nid_list_lock);
2301 update_free_nid_bitmap(sbi, start_nid, false, true);
2302 spin_unlock(&NM_I(sbi)->nid_list_lock);
2303 }
4ac91242 2304 }
e2374015
CY		 2305
2306 return 0;
4ac91242
CY		 2307}
2308
2fbaa25f 2309static void scan_curseg_cache(struct f2fs_sb_info *sbi)
4ac91242 2310{
4ac91242
CY		 2311 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2312 struct f2fs_journal *journal = curseg->journal;
2fbaa25f
CY		 2313 int i;
2314
2315 down_read(&curseg->journal_rwsem);
2316 for (i = 0; i < nats_in_cursum(journal); i++) {
2317 block_t addr;
2318 nid_t nid;
2319
2320 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
2321 nid = le32_to_cpu(nid_in_journal(journal, i));
2322 if (addr == NULL_ADDR)
5921aaa1 2323 add_free_nid(sbi, nid, true, false);
2fbaa25f
CY		 2324 else
2325 remove_free_nid(sbi, nid);
2326 }
2327 up_read(&curseg->journal_rwsem);
2328}
2329
2330static void scan_free_nid_bits(struct f2fs_sb_info *sbi)
2331{
2332 struct f2fs_nm_info *nm_i = NM_I(sbi);
4ac91242 2333 unsigned int i, idx;
97456574 2334 nid_t nid;
4ac91242
CY		 2335
2336 down_read(&nm_i->nat_tree_lock);
2337
2338 for (i = 0; i < nm_i->nat_blocks; i++) {
2339 if (!test_bit_le(i, nm_i->nat_block_bitmap))
2340 continue;
586d1492
CY		 2341 if (!nm_i->free_nid_count[i])
2342 continue;
4ac91242 2343 for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) {
97456574
FL		 2344 idx = find_next_bit_le(nm_i->free_nid_bitmap[i],
2345 NAT_ENTRY_PER_BLOCK, idx);
2346 if (idx >= NAT_ENTRY_PER_BLOCK)
2347 break;
4ac91242
CY		 2348
2349 nid = i * NAT_ENTRY_PER_BLOCK + idx;
5921aaa1 2350 add_free_nid(sbi, nid, true, false);
4ac91242 2351
9a4ffdf5 2352 if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS)
4ac91242
CY		 2353 goto out;
2354 }
2355 }
2356out:
2fbaa25f 2357 scan_curseg_cache(sbi);
4ac91242 2358
4ac91242 2359 up_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 2360}
2361
e2374015 2362static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
4d57b86d 2363 bool sync, bool mount)
e05df3b1 2364{
e05df3b1 2365 struct f2fs_nm_info *nm_i = NM_I(sbi);
e2374015 2366 int i = 0, ret;
55008d84 2367 nid_t nid = nm_i->next_scan_nid;
e05df3b1 2368
e9cdd307
YH		 2369 if (unlikely(nid >= nm_i->max_nid))
2370 nid = 0;
2371
55008d84 2372 /* Enough entries */
9a4ffdf5 2373 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2374 return 0;
e05df3b1 2375
4d57b86d 2376 if (!sync && !f2fs_available_free_memory(sbi, FREE_NIDS))
e2374015 2377 return 0;
e05df3b1 2378
4ac91242
CY		 2379 if (!mount) {
2380 /* try to find free nids in free_nid_bitmap */
2381 scan_free_nid_bits(sbi);
2382
74986213 2383 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2384 return 0;
22ad0b6a
JK		 2385 }
2386
55008d84 2387 /* readahead nat pages to be scanned */
4d57b86d 2388 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
26879fb1 2389 META_NAT, true);
e05df3b1 2390
b873b798 2391 down_read(&nm_i->nat_tree_lock);
a5131193 2392
e05df3b1 2393 while (1) {
66e83361
YH		 2394 if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
2395 nm_i->nat_block_bitmap)) {
2396 struct page *page = get_current_nat_page(sbi, nid);
e05df3b1 2397
edc55aaf
JK		 2398 if (IS_ERR(page)) {
2399 ret = PTR_ERR(page);
2400 } else {
2401 ret = scan_nat_page(sbi, page, nid);
2402 f2fs_put_page(page, 1);
2403 }
e2374015
CY		 2404
2405 if (ret) {
2406 up_read(&nm_i->nat_tree_lock);
dcbb4c10 2407 f2fs_err(sbi, "NAT is corrupt, run fsck to fix it");
edc55aaf 2408 return ret;
e2374015 2409 }
66e83361 2410 }
e05df3b1
JK		 2411
2412 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
cfb271d4 2413 if (unlikely(nid >= nm_i->max_nid))
e05df3b1 2414 nid = 0;
55008d84 2415
a6d494b6 2416 if (++i >= FREE_NID_PAGES)
e05df3b1
JK		 2417 break;
2418 }
2419
55008d84
JK		 2420 /* go to the next free nat pages to find free nids abundantly */
2421 nm_i->next_scan_nid = nid;
e05df3b1
JK		 2422
2423 /* find free nids from current sum_pages */
2fbaa25f 2424 scan_curseg_cache(sbi);
dfc08a12 2425
b873b798 2426 up_read(&nm_i->nat_tree_lock);
2db2388f 2427
4d57b86d 2428 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
ea1a29a0 2429 nm_i->ra_nid_pages, META_NAT, false);
e2374015
CY		 2430
2431 return 0;
e05df3b1
JK		 2432}
2433
e2374015 2434int f2fs_build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
2411cf5b 2435{
e2374015
CY		 2436 int ret;
2437
2411cf5b 2438 mutex_lock(&NM_I(sbi)->build_lock);
e2374015 2439 ret = __f2fs_build_free_nids(sbi, sync, mount);
2411cf5b 2440 mutex_unlock(&NM_I(sbi)->build_lock);
e2374015
CY		 2441
2442 return ret;
2411cf5b
CY		 2443}
2444
e05df3b1
JK		 2445/*
2446 * If this function returns success, caller can obtain a new nid
2447 * from second parameter of this function.
2448 * The returned nid could be used ino as well as nid when inode is created.
2449 */
4d57b86d 2450bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
e05df3b1
JK		 2451{
2452 struct f2fs_nm_info *nm_i = NM_I(sbi);
2453 struct free_nid *i = NULL;
e05df3b1 2454retry:
55523519 2455 if (time_to_inject(sbi, FAULT_ALLOC_NID)) {
c45d6002 2456 f2fs_show_injection_info(sbi, FAULT_ALLOC_NID);
cb78942b 2457 return false;
55523519 2458 }
7fa750a1 2459
b8559dc2 2460 spin_lock(&nm_i->nid_list_lock);
e05df3b1 2461
04d47e67
CY		 2462 if (unlikely(nm_i->available_nids == 0)) {
2463 spin_unlock(&nm_i->nid_list_lock);
2464 return false;
2465 }
e05df3b1 2466
4d57b86d
CY		 2467 /* We should not use stale free nids created by f2fs_build_free_nids */
2468 if (nm_i->nid_cnt[FREE_NID] && !on_f2fs_build_free_nids(nm_i)) {
9a4ffdf5
CY		 2469 f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list));
2470 i = list_first_entry(&nm_i->free_nid_list,
b8559dc2 2471 struct free_nid, list);
55008d84 2472 *nid = i->nid;
b8559dc2 2473
a0761f63 2474 __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID);
04d47e67 2475 nm_i->available_nids--;
4ac91242 2476
346fe752 2477 update_free_nid_bitmap(sbi, *nid, false, false);
4ac91242 2478
b8559dc2 2479 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2480 return true;
2481 }
b8559dc2 2482 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2483
2484 /* Let's scan nat pages and its caches to get free nids */
f84262b0
JK		 2485 if (!f2fs_build_free_nids(sbi, true, false))
2486 goto retry;
2487 return false;
e05df3b1
JK		 2488}
2489
0a8165d7 2490/*
4d57b86d 2491 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2492 */
4d57b86d 2493void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK		 2494{
2495 struct f2fs_nm_info *nm_i = NM_I(sbi);
2496 struct free_nid *i;
2497
b8559dc2 2498 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2499 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2 2500 f2fs_bug_on(sbi, !i);
a0761f63 2501 __remove_free_nid(sbi, i, PREALLOC_NID);
b8559dc2 2502 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2503
2504 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2505}
2506
0a8165d7 2507/*
4d57b86d 2508 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2509 */
4d57b86d 2510void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2511{
49952fa1
JK		 2512 struct f2fs_nm_info *nm_i = NM_I(sbi);
2513 struct free_nid *i;
cf0ee0f0 2514 bool need_free = false;
49952fa1 2515
65985d93
JK		 2516 if (!nid)
2517 return;
2518
b8559dc2 2519 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2520 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2
CY		 2521 f2fs_bug_on(sbi, !i);
2522
4d57b86d 2523 if (!f2fs_available_free_memory(sbi, FREE_NIDS)) {
a0761f63 2524 __remove_free_nid(sbi, i, PREALLOC_NID);
cf0ee0f0 2525 need_free = true;
95630cba 2526 } else {
a0761f63 2527 __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID);
95630cba 2528 }
04d47e67
CY		 2529
2530 nm_i->available_nids++;
2531
346fe752 2532 update_free_nid_bitmap(sbi, nid, true, false);
4ac91242 2533
b8559dc2 2534 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2535
2536 if (need_free)
2537 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2538}
2539
4d57b86d 2540int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
31696580
CY		 2541{
2542 struct f2fs_nm_info *nm_i = NM_I(sbi);
31696580
CY		 2543 int nr = nr_shrink;
2544
9a4ffdf5 2545 if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
ad4edb83
JK		 2546 return 0;
2547
31696580
CY		 2548 if (!mutex_trylock(&nm_i->build_lock))
2549 return 0;
2550
042be373
CY		 2551 while (nr_shrink && nm_i->nid_cnt[FREE_NID] > MAX_FREE_NIDS) {
2552 struct free_nid *i, *next;
2553 unsigned int batch = SHRINK_NID_BATCH_SIZE;
b8559dc2 2554
042be373
CY		 2555 spin_lock(&nm_i->nid_list_lock);
2556 list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) {
2557 if (!nr_shrink || !batch ||
2558 nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
2559 break;
2560 __remove_free_nid(sbi, i, FREE_NID);
2561 kmem_cache_free(free_nid_slab, i);
2562 nr_shrink--;
2563 batch--;
2564 }
2565 spin_unlock(&nm_i->nid_list_lock);
31696580 2566 }
042be373 2567
31696580
CY		 2568 mutex_unlock(&nm_i->build_lock);
2569
2570 return nr - nr_shrink;
2571}
2572
4d57b86d 2573void f2fs_recover_inline_xattr(struct inode *inode, struct page *page)
28cdce04 2574{
28cdce04
CY		 2575 void *src_addr, *dst_addr;
2576 size_t inline_size;
2577 struct page *ipage;
2578 struct f2fs_inode *ri;
2579
4d57b86d 2580 ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino);
9850cf4a 2581 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage));
28cdce04 2582
e3b4d43f 2583 ri = F2FS_INODE(page);
1eca05aa
YH		 2584 if (ri->i_inline & F2FS_INLINE_XATTR) {
2585 set_inode_flag(inode, FI_INLINE_XATTR);
2586 } else {
91942321 2587 clear_inode_flag(inode, FI_INLINE_XATTR);
e3b4d43f
JK		 2588 goto update_inode;
2589 }
2590
6afc662e
CY		 2591 dst_addr = inline_xattr_addr(inode, ipage);
2592 src_addr = inline_xattr_addr(inode, page);
28cdce04
CY		 2593 inline_size = inline_xattr_size(inode);
2594
bae0ee7a 2595 f2fs_wait_on_page_writeback(ipage, NODE, true, true);
28cdce04 2596 memcpy(dst_addr, src_addr, inline_size);
e3b4d43f 2597update_inode:
4d57b86d 2598 f2fs_update_inode(inode, ipage);
28cdce04
CY		 2599 f2fs_put_page(ipage, 1);
2600}
2601
4d57b86d 2602int f2fs_recover_xattr_data(struct inode *inode, struct page *page)
abb2366c 2603{
4081363f 2604 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
abb2366c 2605 nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
87905682
YH		 2606 nid_t new_xnid;
2607 struct dnode_of_data dn;
abb2366c 2608 struct node_info ni;
d260081c 2609 struct page *xpage;
7735730d 2610 int err;
abb2366c 2611
abb2366c
JK		 2612 if (!prev_xnid)
2613 goto recover_xnid;
2614
d260081c 2615 /* 1: invalidate the previous xattr nid */
7735730d
CY		 2616 err = f2fs_get_node_info(sbi, prev_xnid, &ni);
2617 if (err)
2618 return err;
2619
4d57b86d 2620 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 2621 dec_valid_node_count(sbi, inode, false);
479f40c4 2622 set_node_addr(sbi, &ni, NULL_ADDR, false);
abb2366c
JK		 2623
2624recover_xnid:
d260081c 2625 /* 2: update xattr nid in inode */
4d57b86d 2626 if (!f2fs_alloc_nid(sbi, &new_xnid))
87905682
YH		 2627 return -ENOSPC;
2628
2629 set_new_dnode(&dn, inode, NULL, NULL, new_xnid);
4d57b86d 2630 xpage = f2fs_new_node_page(&dn, XATTR_NODE_OFFSET);
87905682 2631 if (IS_ERR(xpage)) {
4d57b86d 2632 f2fs_alloc_nid_failed(sbi, new_xnid);
87905682
YH		 2633 return PTR_ERR(xpage);
2634 }
2635
4d57b86d
CY		 2636 f2fs_alloc_nid_done(sbi, new_xnid);
2637 f2fs_update_inode_page(inode);
d260081c
CY		 2638
2639 /* 3: update and set xattr node page dirty */
87905682 2640 memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE);
abb2366c 2641
d260081c
CY		 2642 set_page_dirty(xpage);
2643 f2fs_put_page(xpage, 1);
abb2366c 2644
d260081c 2645 return 0;
abb2366c
JK		 2646}
2647
4d57b86d 2648int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
e05df3b1 2649{
58bfaf44 2650 struct f2fs_inode *src, *dst;
e05df3b1
JK		 2651 nid_t ino = ino_of_node(page);
2652 struct node_info old_ni, new_ni;
2653 struct page *ipage;
7735730d 2654 int err;
e05df3b1 2655
7735730d
CY		 2656 err = f2fs_get_node_info(sbi, ino, &old_ni);
2657 if (err)
2658 return err;
e8271fa3
JK		 2659
2660 if (unlikely(old_ni.blk_addr != NULL_ADDR))
2661 return -EINVAL;
e8ea9b3d 2662retry:
300e129c 2663 ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false);
e8ea9b3d 2664 if (!ipage) {
5df7731f 2665 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
e8ea9b3d
JK		 2666 goto retry;
2667 }
e05df3b1 2668
e1c42045 2669 /* Should not use this inode from free nid list */
b8559dc2 2670 remove_free_nid(sbi, ino);
e05df3b1 2671
237c0790
JK		 2672 if (!PageUptodate(ipage))
2673 SetPageUptodate(ipage);
e05df3b1 2674 fill_node_footer(ipage, ino, ino, 0, true);
ef2a0071 2675 set_cold_node(ipage, false);
e05df3b1 2676
58bfaf44
JK		 2677 src = F2FS_INODE(page);
2678 dst = F2FS_INODE(ipage);
e05df3b1 2679
58bfaf44
JK		 2680 memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src);
2681 dst->i_size = 0;
2682 dst->i_blocks = cpu_to_le64(1);
2683 dst->i_links = cpu_to_le32(1);
2684 dst->i_xattr_nid = 0;
7a2af766 2685 dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR);
5c57132e 2686 if (dst->i_inline & F2FS_EXTRA_ATTR) {
7a2af766 2687 dst->i_extra_isize = src->i_extra_isize;
6afc662e 2688
7beb01f7 2689 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
6afc662e
CY		 2690 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2691 i_inline_xattr_size))
2692 dst->i_inline_xattr_size = src->i_inline_xattr_size;
2693
7beb01f7 2694 if (f2fs_sb_has_project_quota(sbi) &&
5c57132e
CY		 2695 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2696 i_projid))
2697 dst->i_projid = src->i_projid;
5cd1f387 2698
7beb01f7 2699 if (f2fs_sb_has_inode_crtime(sbi) &&
5cd1f387
CY		 2700 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2701 i_crtime_nsec)) {
2702 dst->i_crtime = src->i_crtime;
2703 dst->i_crtime_nsec = src->i_crtime_nsec;
2704 }
5c57132e 2705 }
e05df3b1
JK		 2706
2707 new_ni = old_ni;
2708 new_ni.ino = ino;
2709
0abd675e 2710 if (unlikely(inc_valid_node_count(sbi, NULL, true)))
65e5cd0a 2711 WARN_ON(1);
479f40c4 2712 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
e05df3b1 2713 inc_valid_inode_count(sbi);
617deb8c 2714 set_page_dirty(ipage);
e05df3b1
JK		 2715 f2fs_put_page(ipage, 1);
2716 return 0;
2717}
2718
7735730d 2719int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
e05df3b1
JK		 2720 unsigned int segno, struct f2fs_summary_block *sum)
2721{
2722 struct f2fs_node *rn;
2723 struct f2fs_summary *sum_entry;
e05df3b1 2724 block_t addr;
9ecf4b80 2725 int i, idx, last_offset, nrpages;
e05df3b1
JK		 2726
2727 /* scan the node segment */
2728 last_offset = sbi->blocks_per_seg;
2729 addr = START_BLOCK(sbi, segno);
2730 sum_entry = &sum->entries[0];
2731
9ecf4b80 2732 for (i = 0; i < last_offset; i += nrpages, addr += nrpages) {
664ba972 2733 nrpages = min(last_offset - i, BIO_MAX_PAGES);
393ff91f 2734
e1c42045 2735 /* readahead node pages */
4d57b86d 2736 f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true);
e05df3b1 2737
9ecf4b80 2738 for (idx = addr; idx < addr + nrpages; idx++) {
4d57b86d 2739 struct page *page = f2fs_get_tmp_page(sbi, idx);
9af0ff1c 2740
7735730d
CY		 2741 if (IS_ERR(page))
2742 return PTR_ERR(page);
2743
9ecf4b80
CY		 2744 rn = F2FS_NODE(page);
2745 sum_entry->nid = rn->footer.nid;
2746 sum_entry->version = 0;
2747 sum_entry->ofs_in_node = 0;
2748 sum_entry++;
2749 f2fs_put_page(page, 1);
9af0ff1c 2750 }
bac4eef6 2751
9ecf4b80 2752 invalidate_mapping_pages(META_MAPPING(sbi), addr,
bac4eef6 2753 addr + nrpages);
e05df3b1 2754 }
7735730d 2755 return 0;
e05df3b1
JK		 2756}
2757
aec71382 2758static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
e05df3b1
JK		 2759{
2760 struct f2fs_nm_info *nm_i = NM_I(sbi);
2761 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2762 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 2763 int i;
2764
b7ad7512 2765 down_write(&curseg->journal_rwsem);
dfc08a12 2766 for (i = 0; i < nats_in_cursum(journal); i++) {
e05df3b1
JK		 2767 struct nat_entry *ne;
2768 struct f2fs_nat_entry raw_ne;
dfc08a12 2769 nid_t nid = le32_to_cpu(nid_in_journal(journal, i));
e05df3b1 2770
dfc08a12 2771 raw_ne = nat_in_journal(journal, i);
9be32d72 2772
e05df3b1 2773 ne = __lookup_nat_cache(nm_i, nid);
e05df3b1 2774 if (!ne) {
12f9ef37
YH		 2775 ne = __alloc_nat_entry(nid, true);
2776 __init_nat_entry(nm_i, ne, &raw_ne, true);
e05df3b1 2777 }
04d47e67
CY		 2778
2779 /*
2780 * if a free nat in journal has not been used after last
2781 * checkpoint, we should remove it from available nids,
2782 * since later we will add it again.
2783 */
2784 if (!get_nat_flag(ne, IS_DIRTY) &&
2785 le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) {
2786 spin_lock(&nm_i->nid_list_lock);
2787 nm_i->available_nids--;
2788 spin_unlock(&nm_i->nid_list_lock);
2789 }
2790
e05df3b1 2791 __set_nat_cache_dirty(nm_i, ne);
e05df3b1 2792 }
dfc08a12 2793 update_nats_in_cursum(journal, -i);
b7ad7512 2794 up_write(&curseg->journal_rwsem);
e05df3b1
JK		 2795}
2796
309cc2b6
JK		 2797static void __adjust_nat_entry_set(struct nat_entry_set *nes,
2798 struct list_head *head, int max)
e05df3b1 2799{
309cc2b6 2800 struct nat_entry_set *cur;
e05df3b1 2801
309cc2b6
JK		 2802 if (nes->entry_cnt >= max)
2803 goto add_out;
e05df3b1 2804
309cc2b6
JK		 2805 list_for_each_entry(cur, head, set_list) {
2806 if (cur->entry_cnt >= nes->entry_cnt) {
2807 list_add(&nes->set_list, cur->set_list.prev);
2808 return;
2809 }
aec71382 2810 }
309cc2b6
JK		 2811add_out:
2812 list_add_tail(&nes->set_list, head);
2813}
e05df3b1 2814
9f7e4a2c 2815static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
22ad0b6a
JK		 2816 struct page *page)
2817{
2818 struct f2fs_nm_info *nm_i = NM_I(sbi);
2819 unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK;
2820 struct f2fs_nat_block *nat_blk = page_address(page);
2821 int valid = 0;
37a0ab2a 2822 int i = 0;
22ad0b6a
JK		 2823
2824 if (!enabled_nat_bits(sbi, NULL))
2825 return;
2826
37a0ab2a
FL		 2827 if (nat_index == 0) {
2828 valid = 1;
2829 i = 1;
2830 }
2831 for (; i < NAT_ENTRY_PER_BLOCK; i++) {
36af5f40 2832 if (le32_to_cpu(nat_blk->entries[i].block_addr) != NULL_ADDR)
22ad0b6a
JK		 2833 valid++;
2834 }
2835 if (valid == 0) {
23380b85
JK		 2836 __set_bit_le(nat_index, nm_i->empty_nat_bits);
2837 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK		 2838 return;
2839 }
2840
23380b85 2841 __clear_bit_le(nat_index, nm_i->empty_nat_bits);
22ad0b6a 2842 if (valid == NAT_ENTRY_PER_BLOCK)
23380b85 2843 __set_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a 2844 else
23380b85 2845 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK		 2846}
2847
edc55aaf 2848static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
22ad0b6a 2849 struct nat_entry_set *set, struct cp_control *cpc)
309cc2b6
JK		 2850{
2851 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2852 struct f2fs_journal *journal = curseg->journal;
309cc2b6
JK		 2853 nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK;
2854 bool to_journal = true;
2855 struct f2fs_nat_block *nat_blk;
2856 struct nat_entry *ne, *cur;
2857 struct page *page = NULL;
e05df3b1 2858
aec71382
CY		 2859 /*
2860 * there are two steps to flush nat entries:
2861 * #1, flush nat entries to journal in current hot data summary block.
2862 * #2, flush nat entries to nat page.
2863 */
22ad0b6a
JK		 2864 if (enabled_nat_bits(sbi, cpc) ||
2865 !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL))
309cc2b6
JK		 2866 to_journal = false;
2867
2868 if (to_journal) {
b7ad7512 2869 down_write(&curseg->journal_rwsem);
309cc2b6
JK		 2870 } else {
2871 page = get_next_nat_page(sbi, start_nid);
edc55aaf
JK		 2872 if (IS_ERR(page))
2873 return PTR_ERR(page);
2874
309cc2b6
JK		 2875 nat_blk = page_address(page);
2876 f2fs_bug_on(sbi, !nat_blk);
2877 }
aec71382 2878
309cc2b6
JK		 2879 /* flush dirty nats in nat entry set */
2880 list_for_each_entry_safe(ne, cur, &set->entry_list, list) {
2881 struct f2fs_nat_entry *raw_ne;
2882 nid_t nid = nat_get_nid(ne);
2883 int offset;
2884
febeca6d 2885 f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR);
aec71382
CY		 2886
2887 if (to_journal) {
4d57b86d 2888 offset = f2fs_lookup_journal_in_cursum(journal,
309cc2b6
JK		 2889 NAT_JOURNAL, nid, 1);
2890 f2fs_bug_on(sbi, offset < 0);
dfc08a12
CY		 2891 raw_ne = &nat_in_journal(journal, offset);
2892 nid_in_journal(journal, offset) = cpu_to_le32(nid);
aec71382 2893 } else {
309cc2b6 2894 raw_ne = &nat_blk->entries[nid - start_nid];
e05df3b1 2895 }
309cc2b6 2896 raw_nat_from_node_info(raw_ne, &ne->ni);
309cc2b6 2897 nat_reset_flag(ne);
0b28b71e 2898 __clear_nat_cache_dirty(NM_I(sbi), set, ne);
04d47e67 2899 if (nat_get_blkaddr(ne) == NULL_ADDR) {
5921aaa1 2900 add_free_nid(sbi, nid, false, true);
4ac91242
CY		 2901 } else {
2902 spin_lock(&NM_I(sbi)->nid_list_lock);
346fe752 2903 update_free_nid_bitmap(sbi, nid, false, false);
04d47e67
CY		 2904 spin_unlock(&NM_I(sbi)->nid_list_lock);
2905 }
309cc2b6 2906 }
e05df3b1 2907
22ad0b6a 2908 if (to_journal) {
b7ad7512 2909 up_write(&curseg->journal_rwsem);
22ad0b6a
JK		 2910 } else {
2911 __update_nat_bits(sbi, start_nid, page);
309cc2b6 2912 f2fs_put_page(page, 1);
22ad0b6a 2913 }
aec71382 2914
59c9081b
YH		 2915 /* Allow dirty nats by node block allocation in write_begin */
2916 if (!set->entry_cnt) {
2917 radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
2918 kmem_cache_free(nat_entry_set_slab, set);
2919 }
edc55aaf 2920 return 0;
309cc2b6 2921}
aec71382 2922
309cc2b6
JK		 2923/*
2924 * This function is called during the checkpointing process.
2925 */
edc55aaf 2926int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
309cc2b6
JK		 2927{
2928 struct f2fs_nm_info *nm_i = NM_I(sbi);
2929 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2930 struct f2fs_journal *journal = curseg->journal;
7aed0d45 2931 struct nat_entry_set *setvec[SETVEC_SIZE];
309cc2b6
JK		 2932 struct nat_entry_set *set, *tmp;
2933 unsigned int found;
2934 nid_t set_idx = 0;
2935 LIST_HEAD(sets);
edc55aaf 2936 int err = 0;
309cc2b6 2937
7f2ecdd8
JK		 2938 /* during unmount, let's flush nat_bits before checking dirty_nat_cnt */
2939 if (enabled_nat_bits(sbi, cpc)) {
2940 down_write(&nm_i->nat_tree_lock);
2941 remove_nats_in_journal(sbi);
2942 up_write(&nm_i->nat_tree_lock);
2943 }
2944
20d047c8 2945 if (!nm_i->dirty_nat_cnt)
edc55aaf 2946 return 0;
a5131193 2947
b873b798 2948 down_write(&nm_i->nat_tree_lock);
a5131193 2949
309cc2b6
JK		 2950 /*
2951 * if there are no enough space in journal to store dirty nat
2952 * entries, remove all entries from journal and merge them
2953 * into nat entry set.
2954 */
900f7362 2955 if (enabled_nat_bits(sbi, cpc) ||
22ad0b6a 2956 !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL))
309cc2b6
JK		 2957 remove_nats_in_journal(sbi);
2958
309cc2b6 2959 while ((found = __gang_lookup_nat_set(nm_i,
7aed0d45 2960 set_idx, SETVEC_SIZE, setvec))) {
309cc2b6
JK		 2961 unsigned idx;
2962 set_idx = setvec[found - 1]->set + 1;
2963 for (idx = 0; idx < found; idx++)
2964 __adjust_nat_entry_set(setvec[idx], &sets,
dfc08a12 2965 MAX_NAT_JENTRIES(journal));
e05df3b1 2966 }
aec71382 2967
309cc2b6 2968 /* flush dirty nats in nat entry set */
edc55aaf
JK		 2969 list_for_each_entry_safe(set, tmp, &sets, set_list) {
2970 err = __flush_nat_entry_set(sbi, set, cpc);
2971 if (err)
2972 break;
2973 }
309cc2b6 2974
b873b798 2975 up_write(&nm_i->nat_tree_lock);
59c9081b 2976 /* Allow dirty nats by node block allocation in write_begin */
edc55aaf
JK		 2977
2978 return err;
e05df3b1
JK		 2979}
2980
22ad0b6a
JK		 2981static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
2982{
2983 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2984 struct f2fs_nm_info *nm_i = NM_I(sbi);
2985 unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE;
2986 unsigned int i;
2987 __u64 cp_ver = cur_cp_version(ckpt);
22ad0b6a
JK		 2988 block_t nat_bits_addr;
2989
2990 if (!enabled_nat_bits(sbi, NULL))
2991 return 0;
2992
df033caf 2993 nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
0b6d4ca0 2994 nm_i->nat_bits = f2fs_kvzalloc(sbi,
acbf054d 2995 nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL);
22ad0b6a
JK		 2996 if (!nm_i->nat_bits)
2997 return -ENOMEM;
2998
2999 nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg -
3000 nm_i->nat_bits_blocks;
3001 for (i = 0; i < nm_i->nat_bits_blocks; i++) {
7735730d
CY		 3002 struct page *page;
3003
3004 page = f2fs_get_meta_page(sbi, nat_bits_addr++);
3b30eb19 3005 if (IS_ERR(page))
7735730d 3006 return PTR_ERR(page);
22ad0b6a
JK		 3007
3008 memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS),
3009 page_address(page), F2FS_BLKSIZE);
3010 f2fs_put_page(page, 1);
3011 }
3012
ced2c7ea 3013 cp_ver |= (cur_cp_crc(ckpt) << 32);
22ad0b6a
JK		 3014 if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) {
3015 disable_nat_bits(sbi, true);
3016 return 0;
3017 }
3018
3019 nm_i->full_nat_bits = nm_i->nat_bits + 8;
3020 nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes;
3021
dcbb4c10 3022 f2fs_notice(sbi, "Found nat_bits in checkpoint");
22ad0b6a
JK		 3023 return 0;
3024}
3025
bd80a4b9 3026static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi)
7041d5d2
CY		 3027{
3028 struct f2fs_nm_info *nm_i = NM_I(sbi);
3029 unsigned int i = 0;
3030 nid_t nid, last_nid;
3031
3032 if (!enabled_nat_bits(sbi, NULL))
3033 return;
3034
3035 for (i = 0; i < nm_i->nat_blocks; i++) {
3036 i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i);
3037 if (i >= nm_i->nat_blocks)
3038 break;
3039
3040 __set_bit_le(i, nm_i->nat_block_bitmap);
3041
3042 nid = i * NAT_ENTRY_PER_BLOCK;
f6986ede 3043 last_nid = nid + NAT_ENTRY_PER_BLOCK;
7041d5d2 3044
346fe752 3045 spin_lock(&NM_I(sbi)->nid_list_lock);
7041d5d2 3046 for (; nid < last_nid; nid++)
346fe752
CY		 3047 update_free_nid_bitmap(sbi, nid, true, true);
3048 spin_unlock(&NM_I(sbi)->nid_list_lock);
7041d5d2
CY		 3049 }
3050
3051 for (i = 0; i < nm_i->nat_blocks; i++) {
3052 i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i);
3053 if (i >= nm_i->nat_blocks)
3054 break;
3055
3056 __set_bit_le(i, nm_i->nat_block_bitmap);
3057 }
3058}
3059
e05df3b1
JK		 3060static int init_node_manager(struct f2fs_sb_info *sbi)
3061{
3062 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
3063 struct f2fs_nm_info *nm_i = NM_I(sbi);
3064 unsigned char *version_bitmap;
22ad0b6a
JK		 3065 unsigned int nat_segs;
3066 int err;
e05df3b1
JK		 3067
3068 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
3069
3070 /* segment_count_nat includes pair segment so divide to 2. */
3071 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
22ad0b6a
JK		 3072 nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
3073 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
7ee0eeab 3074
b63da15e 3075 /* not used nids: 0, node, meta, (and root counted as valid node) */
04d47e67 3076 nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count -
27cae0bc 3077 F2FS_RESERVED_NODE_NUM;
9a4ffdf5
CY		 3078 nm_i->nid_cnt[FREE_NID] = 0;
3079 nm_i->nid_cnt[PREALLOC_NID] = 0;
e05df3b1 3080 nm_i->nat_cnt = 0;
cdfc41c1 3081 nm_i->ram_thresh = DEF_RAM_THRESHOLD;
ea1a29a0 3082 nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
2304cb0c 3083 nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
e05df3b1 3084
8a7ed66a 3085 INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
9a4ffdf5 3086 INIT_LIST_HEAD(&nm_i->free_nid_list);
769ec6e5
JK		 3087 INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO);
3088 INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO);
e05df3b1 3089 INIT_LIST_HEAD(&nm_i->nat_entries);
22969158 3090 spin_lock_init(&nm_i->nat_list_lock);
e05df3b1
JK		 3091
3092 mutex_init(&nm_i->build_lock);
b8559dc2 3093 spin_lock_init(&nm_i->nid_list_lock);
b873b798 3094 init_rwsem(&nm_i->nat_tree_lock);
e05df3b1 3095
e05df3b1 3096 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
79b5793b 3097 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
e05df3b1
JK		 3098 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
3099 if (!version_bitmap)
3100 return -EFAULT;
3101
79b5793b
AG		 3102 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
3103 GFP_KERNEL);
3104 if (!nm_i->nat_bitmap)
3105 return -ENOMEM;
599a09b2 3106
22ad0b6a
JK		 3107 err = __get_nat_bitmaps(sbi);
3108 if (err)
3109 return err;
3110
599a09b2
CY		 3111#ifdef CONFIG_F2FS_CHECK_FS
3112 nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size,
3113 GFP_KERNEL);
3114 if (!nm_i->nat_bitmap_mir)
3115 return -ENOMEM;
3116#endif
3117
e05df3b1
JK		 3118 return 0;
3119}
3120
9f7e4a2c 3121static int init_free_nid_cache(struct f2fs_sb_info *sbi)
4ac91242
CY		 3122{
3123 struct f2fs_nm_info *nm_i = NM_I(sbi);
bb1105e4 3124 int i;
4ac91242 3125
026f0507 3126 nm_i->free_nid_bitmap =
0b6d4ca0
EB		 3127 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned char *),
3128 nm_i->nat_blocks),
3129 GFP_KERNEL);
4ac91242
CY		 3130 if (!nm_i->free_nid_bitmap)
3131 return -ENOMEM;
3132
bb1105e4
JK		 3133 for (i = 0; i < nm_i->nat_blocks; i++) {
3134 nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi,
e15d54d5 3135 f2fs_bitmap_size(NAT_ENTRY_PER_BLOCK), GFP_KERNEL);
68c43a23 3136 if (!nm_i->free_nid_bitmap[i])
bb1105e4
JK		 3137 return -ENOMEM;
3138 }
3139
628b3d14 3140 nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8,
4ac91242
CY		 3141 GFP_KERNEL);
3142 if (!nm_i->nat_block_bitmap)
3143 return -ENOMEM;
586d1492 3144
9d2a789c
KC		 3145 nm_i->free_nid_count =
3146 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned short),
3147 nm_i->nat_blocks),
3148 GFP_KERNEL);
586d1492
CY		 3149 if (!nm_i->free_nid_count)
3150 return -ENOMEM;
4ac91242
CY		 3151 return 0;
3152}
3153
4d57b86d 3154int f2fs_build_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK		 3155{
3156 int err;
3157
acbf054d
CY		 3158 sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info),
3159 GFP_KERNEL);
e05df3b1
JK		 3160 if (!sbi->nm_info)
3161 return -ENOMEM;
3162
3163 err = init_node_manager(sbi);
3164 if (err)
3165 return err;
3166
4ac91242
CY		 3167 err = init_free_nid_cache(sbi);
3168 if (err)
3169 return err;
3170
7041d5d2
CY		 3171 /* load free nid status from nat_bits table */
3172 load_free_nid_bitmap(sbi);
3173
e2374015 3174 return f2fs_build_free_nids(sbi, true, true);
e05df3b1
JK		 3175}
3176
4d57b86d 3177void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK		 3178{
3179 struct f2fs_nm_info *nm_i = NM_I(sbi);
3180 struct free_nid *i, *next_i;
3181 struct nat_entry *natvec[NATVEC_SIZE];
7aed0d45 3182 struct nat_entry_set *setvec[SETVEC_SIZE];
e05df3b1
JK		 3183 nid_t nid = 0;
3184 unsigned int found;
3185
3186 if (!nm_i)
3187 return;
3188
3189 /* destroy free nid list */
b8559dc2 3190 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 3191 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
a0761f63 3192 __remove_free_nid(sbi, i, FREE_NID);
b8559dc2 3193 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0 3194 kmem_cache_free(free_nid_slab, i);
b8559dc2 3195 spin_lock(&nm_i->nid_list_lock);
e05df3b1 3196 }
9a4ffdf5
CY		 3197 f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]);
3198 f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]);
3199 f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list));
b8559dc2 3200 spin_unlock(&nm_i->nid_list_lock);
e05df3b1
JK		 3201
3202 /* destroy nat cache */
b873b798 3203 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 3204 while ((found = __gang_lookup_nat_cache(nm_i,
3205 nid, NATVEC_SIZE, natvec))) {
3206 unsigned idx;
7aed0d45 3207
b6ce391e 3208 nid = nat_get_nid(natvec[found - 1]) + 1;
22969158
CY		 3209 for (idx = 0; idx < found; idx++) {
3210 spin_lock(&nm_i->nat_list_lock);
3211 list_del(&natvec[idx]->list);
3212 spin_unlock(&nm_i->nat_list_lock);
3213
b6ce391e 3214 __del_from_nat_cache(nm_i, natvec[idx]);
22969158 3215 }
e05df3b1 3216 }
9850cf4a 3217 f2fs_bug_on(sbi, nm_i->nat_cnt);
7aed0d45
JK		 3218
3219 /* destroy nat set cache */
3220 nid = 0;
3221 while ((found = __gang_lookup_nat_set(nm_i,
3222 nid, SETVEC_SIZE, setvec))) {
3223 unsigned idx;
3224
3225 nid = setvec[found - 1]->set + 1;
3226 for (idx = 0; idx < found; idx++) {
3227 /* entry_cnt is not zero, when cp_error was occurred */
3228 f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list));
3229 radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set);
3230 kmem_cache_free(nat_entry_set_slab, setvec[idx]);
3231 }
3232 }
b873b798 3233 up_write(&nm_i->nat_tree_lock);
e05df3b1 3234
4ac91242 3235 kvfree(nm_i->nat_block_bitmap);
bb1105e4
JK		 3236 if (nm_i->free_nid_bitmap) {
3237 int i;
3238
3239 for (i = 0; i < nm_i->nat_blocks; i++)
3240 kvfree(nm_i->free_nid_bitmap[i]);
5222595d 3241 kvfree(nm_i->free_nid_bitmap);
bb1105e4 3242 }
586d1492 3243 kvfree(nm_i->free_nid_count);
4ac91242 3244
5222595d
JK		 3245 kvfree(nm_i->nat_bitmap);
3246 kvfree(nm_i->nat_bits);
599a09b2 3247#ifdef CONFIG_F2FS_CHECK_FS
5222595d 3248 kvfree(nm_i->nat_bitmap_mir);
599a09b2 3249#endif
e05df3b1 3250 sbi->nm_info = NULL;
5222595d 3251 kvfree(nm_i);
e05df3b1
JK		 3252}
3253
4d57b86d 3254int __init f2fs_create_node_manager_caches(void)
e05df3b1 3255{
98510003 3256 nat_entry_slab = f2fs_kmem_cache_create("f2fs_nat_entry",
e8512d2e 3257 sizeof(struct nat_entry));
e05df3b1 3258 if (!nat_entry_slab)
aec71382 3259 goto fail;
e05df3b1 3260
98510003 3261 free_nid_slab = f2fs_kmem_cache_create("f2fs_free_nid",
e8512d2e 3262 sizeof(struct free_nid));
aec71382 3263 if (!free_nid_slab)
ce3e6d25 3264 goto destroy_nat_entry;
aec71382 3265
98510003 3266 nat_entry_set_slab = f2fs_kmem_cache_create("f2fs_nat_entry_set",
aec71382
CY		 3267 sizeof(struct nat_entry_set));
3268 if (!nat_entry_set_slab)
ce3e6d25 3269 goto destroy_free_nid;
50fa53ec 3270
98510003 3271 fsync_node_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_node_entry",
50fa53ec
CY		 3272 sizeof(struct fsync_node_entry));
3273 if (!fsync_node_entry_slab)
3274 goto destroy_nat_entry_set;
e05df3b1 3275 return 0;
aec71382 3276
50fa53ec
CY		 3277destroy_nat_entry_set:
3278 kmem_cache_destroy(nat_entry_set_slab);
ce3e6d25 3279destroy_free_nid:
aec71382 3280 kmem_cache_destroy(free_nid_slab);
ce3e6d25 3281destroy_nat_entry:
aec71382
CY		 3282 kmem_cache_destroy(nat_entry_slab);
3283fail:
3284 return -ENOMEM;
e05df3b1
JK		 3285}
3286
4d57b86d 3287void f2fs_destroy_node_manager_caches(void)
e05df3b1 3288{
50fa53ec 3289 kmem_cache_destroy(fsync_node_entry_slab);
aec71382 3290 kmem_cache_destroy(nat_entry_set_slab);
e05df3b1
JK		 3291 kmem_cache_destroy(free_nid_slab);
3292 kmem_cache_destroy(nat_entry_slab);
3293}
Linux 6.x block layer and io_uring tree(s)