]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - fs/f2fs/node.c
f2fs: fix validation of the block count in sanity_check_raw_super
[mirror_ubuntu-jammy-kernel.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);
37 f2fs_msg(sbi->sb, KERN_WARNING,
38 "%s: out-of-range nid=%x, run fsck to fix.",
39 __func__, nid);
40 return -EINVAL;
41 }
42 return 0;
43}
44
4d57b86d 45bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type)
cdfc41c1 46{
6fb03f3a 47 struct f2fs_nm_info *nm_i = NM_I(sbi);
cdfc41c1 48 struct sysinfo val;
e5e7ea3c 49 unsigned long avail_ram;
cdfc41c1 50 unsigned long mem_size = 0;
6fb03f3a 51 bool res = false;
cdfc41c1
JK
52
53 si_meminfo(&val);
e5e7ea3c
JK
54
55 /* only uses low memory */
56 avail_ram = val.totalram - val.totalhigh;
57
429511cd
CY
58 /*
59 * give 25%, 25%, 50%, 50%, 50% memory for each components respectively
60 */
6fb03f3a 61 if (type == FREE_NIDS) {
9a4ffdf5 62 mem_size = (nm_i->nid_cnt[FREE_NID] *
b8559dc2 63 sizeof(struct free_nid)) >> PAGE_SHIFT;
e5e7ea3c 64 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
6fb03f3a 65 } else if (type == NAT_ENTRIES) {
e5e7ea3c 66 mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >>
09cbfeaf 67 PAGE_SHIFT;
e5e7ea3c 68 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
e589c2c4
JK
69 if (excess_cached_nats(sbi))
70 res = false;
a1257023
JK
71 } else if (type == DIRTY_DENTS) {
72 if (sbi->sb->s_bdi->wb.dirty_exceeded)
73 return false;
74 mem_size = get_pages(sbi, F2FS_DIRTY_DENTS);
75 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
e5e7ea3c
JK
76 } else if (type == INO_ENTRIES) {
77 int i;
78
39d787be 79 for (i = 0; i < MAX_INO_ENTRY; i++)
8f73cbb7
KM
80 mem_size += sbi->im[i].ino_num *
81 sizeof(struct ino_entry);
82 mem_size >>= PAGE_SHIFT;
e5e7ea3c 83 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
429511cd 84 } else if (type == EXTENT_CACHE) {
7441ccef
JK
85 mem_size = (atomic_read(&sbi->total_ext_tree) *
86 sizeof(struct extent_tree) +
429511cd 87 atomic_read(&sbi->total_ext_node) *
09cbfeaf 88 sizeof(struct extent_node)) >> PAGE_SHIFT;
429511cd 89 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
57864ae5
JK
90 } else if (type == INMEM_PAGES) {
91 /* it allows 20% / total_ram for inmemory pages */
92 mem_size = get_pages(sbi, F2FS_INMEM_PAGES);
93 res = mem_size < (val.totalram / 5);
1e84371f 94 } else {
1663cae4
JK
95 if (!sbi->sb->s_bdi->wb.dirty_exceeded)
96 return true;
6fb03f3a
JK
97 }
98 return res;
cdfc41c1
JK
99}
100
e05df3b1
JK
101static void clear_node_page_dirty(struct page *page)
102{
e05df3b1 103 if (PageDirty(page)) {
5ec2d99d 104 f2fs_clear_page_cache_dirty_tag(page);
e05df3b1 105 clear_page_dirty_for_io(page);
aec2f729 106 dec_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
e05df3b1
JK
107 }
108 ClearPageUptodate(page);
109}
110
111static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
112{
80551d17 113 return f2fs_get_meta_page_nofail(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK
114}
115
116static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
117{
118 struct page *src_page;
119 struct page *dst_page;
e05df3b1
JK
120 pgoff_t dst_off;
121 void *src_addr;
122 void *dst_addr;
123 struct f2fs_nm_info *nm_i = NM_I(sbi);
124
80551d17 125 dst_off = next_nat_addr(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK
126
127 /* get current nat block page with lock */
80551d17 128 src_page = get_current_nat_page(sbi, nid);
edc55aaf
JK
129 if (IS_ERR(src_page))
130 return src_page;
4d57b86d 131 dst_page = f2fs_grab_meta_page(sbi, dst_off);
9850cf4a 132 f2fs_bug_on(sbi, PageDirty(src_page));
e05df3b1
JK
133
134 src_addr = page_address(src_page);
135 dst_addr = page_address(dst_page);
09cbfeaf 136 memcpy(dst_addr, src_addr, PAGE_SIZE);
e05df3b1
JK
137 set_page_dirty(dst_page);
138 f2fs_put_page(src_page, 1);
139
140 set_to_next_nat(nm_i, nid);
141
142 return dst_page;
143}
144
12f9ef37
YH
145static struct nat_entry *__alloc_nat_entry(nid_t nid, bool no_fail)
146{
147 struct nat_entry *new;
148
149 if (no_fail)
2882d343 150 new = f2fs_kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37 151 else
2882d343 152 new = kmem_cache_alloc(nat_entry_slab, GFP_F2FS_ZERO);
12f9ef37
YH
153 if (new) {
154 nat_set_nid(new, nid);
155 nat_reset_flag(new);
156 }
157 return new;
158}
159
160static void __free_nat_entry(struct nat_entry *e)
161{
162 kmem_cache_free(nat_entry_slab, e);
163}
164
165/* must be locked by nat_tree_lock */
166static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
167 struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail)
168{
169 if (no_fail)
170 f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne);
171 else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne))
172 return NULL;
173
174 if (raw_ne)
175 node_info_from_raw_nat(&ne->ni, raw_ne);
22969158
CY
176
177 spin_lock(&nm_i->nat_list_lock);
12f9ef37 178 list_add_tail(&ne->list, &nm_i->nat_entries);
22969158
CY
179 spin_unlock(&nm_i->nat_list_lock);
180
12f9ef37
YH
181 nm_i->nat_cnt++;
182 return ne;
183}
184
e05df3b1
JK
185static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
186{
22969158
CY
187 struct nat_entry *ne;
188
189 ne = radix_tree_lookup(&nm_i->nat_root, n);
190
191 /* for recent accessed nat entry, move it to tail of lru list */
192 if (ne && !get_nat_flag(ne, IS_DIRTY)) {
193 spin_lock(&nm_i->nat_list_lock);
194 if (!list_empty(&ne->list))
195 list_move_tail(&ne->list, &nm_i->nat_entries);
196 spin_unlock(&nm_i->nat_list_lock);
197 }
198
199 return ne;
e05df3b1
JK
200}
201
202static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
203 nid_t start, unsigned int nr, struct nat_entry **ep)
204{
205 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
206}
207
208static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
209{
e05df3b1
JK
210 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
211 nm_i->nat_cnt--;
12f9ef37 212 __free_nat_entry(e);
e05df3b1
JK
213}
214
780de47c
CY
215static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
216 struct nat_entry *ne)
309cc2b6
JK
217{
218 nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid);
219 struct nat_entry_set *head;
220
309cc2b6
JK
221 head = radix_tree_lookup(&nm_i->nat_set_root, set);
222 if (!head) {
80c54505 223 head = f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_NOFS);
309cc2b6
JK
224
225 INIT_LIST_HEAD(&head->entry_list);
226 INIT_LIST_HEAD(&head->set_list);
227 head->set = set;
228 head->entry_cnt = 0;
9be32d72 229 f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
309cc2b6 230 }
780de47c
CY
231 return head;
232}
233
234static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
235 struct nat_entry *ne)
236{
237 struct nat_entry_set *head;
238 bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR;
239
240 if (!new_ne)
241 head = __grab_nat_entry_set(nm_i, ne);
242
243 /*
244 * update entry_cnt in below condition:
245 * 1. update NEW_ADDR to valid block address;
246 * 2. update old block address to new one;
247 */
248 if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
249 !get_nat_flag(ne, IS_DIRTY)))
250 head->entry_cnt++;
251
252 set_nat_flag(ne, IS_PREALLOC, new_ne);
febeca6d
CY
253
254 if (get_nat_flag(ne, IS_DIRTY))
255 goto refresh_list;
256
309cc2b6 257 nm_i->dirty_nat_cnt++;
309cc2b6 258 set_nat_flag(ne, IS_DIRTY, true);
febeca6d 259refresh_list:
22969158 260 spin_lock(&nm_i->nat_list_lock);
780de47c 261 if (new_ne)
febeca6d
CY
262 list_del_init(&ne->list);
263 else
264 list_move_tail(&ne->list, &head->entry_list);
22969158 265 spin_unlock(&nm_i->nat_list_lock);
309cc2b6
JK
266}
267
268static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
0b28b71e 269 struct nat_entry_set *set, struct nat_entry *ne)
309cc2b6 270{
22969158 271 spin_lock(&nm_i->nat_list_lock);
0b28b71e 272 list_move_tail(&ne->list, &nm_i->nat_entries);
22969158
CY
273 spin_unlock(&nm_i->nat_list_lock);
274
0b28b71e
KM
275 set_nat_flag(ne, IS_DIRTY, false);
276 set->entry_cnt--;
277 nm_i->dirty_nat_cnt--;
309cc2b6
JK
278}
279
280static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
281 nid_t start, unsigned int nr, struct nat_entry_set **ep)
282{
283 return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep,
284 start, nr);
285}
286
50fa53ec
CY
287bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page)
288{
289 return NODE_MAPPING(sbi) == page->mapping &&
290 IS_DNODE(page) && is_cold_node(page);
291}
292
293void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi)
294{
295 spin_lock_init(&sbi->fsync_node_lock);
296 INIT_LIST_HEAD(&sbi->fsync_node_list);
297 sbi->fsync_seg_id = 0;
298 sbi->fsync_node_num = 0;
299}
300
301static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi,
302 struct page *page)
303{
304 struct fsync_node_entry *fn;
305 unsigned long flags;
306 unsigned int seq_id;
307
308 fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, GFP_NOFS);
309
310 get_page(page);
311 fn->page = page;
312 INIT_LIST_HEAD(&fn->list);
313
314 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
315 list_add_tail(&fn->list, &sbi->fsync_node_list);
316 fn->seq_id = sbi->fsync_seg_id++;
317 seq_id = fn->seq_id;
318 sbi->fsync_node_num++;
319 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
320
321 return seq_id;
322}
323
324void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page)
325{
326 struct fsync_node_entry *fn;
327 unsigned long flags;
328
329 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
330 list_for_each_entry(fn, &sbi->fsync_node_list, list) {
331 if (fn->page == page) {
332 list_del(&fn->list);
333 sbi->fsync_node_num--;
334 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
335 kmem_cache_free(fsync_node_entry_slab, fn);
336 put_page(page);
337 return;
338 }
339 }
340 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
341 f2fs_bug_on(sbi, 1);
342}
343
344void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi)
345{
346 unsigned long flags;
347
348 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
349 sbi->fsync_seg_id = 0;
350 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
351}
352
4d57b86d 353int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK
354{
355 struct f2fs_nm_info *nm_i = NM_I(sbi);
356 struct nat_entry *e;
2dcf51ab 357 bool need = false;
e05df3b1 358
b873b798 359 down_read(&nm_i->nat_tree_lock);
e05df3b1 360 e = __lookup_nat_cache(nm_i, nid);
2dcf51ab
JK
361 if (e) {
362 if (!get_nat_flag(e, IS_CHECKPOINTED) &&
363 !get_nat_flag(e, HAS_FSYNCED_INODE))
364 need = true;
365 }
b873b798 366 up_read(&nm_i->nat_tree_lock);
2dcf51ab 367 return need;
e05df3b1
JK
368}
369
4d57b86d 370bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
479f40c4
JK
371{
372 struct f2fs_nm_info *nm_i = NM_I(sbi);
373 struct nat_entry *e;
2dcf51ab 374 bool is_cp = true;
479f40c4 375
b873b798 376 down_read(&nm_i->nat_tree_lock);
2dcf51ab
JK
377 e = __lookup_nat_cache(nm_i, nid);
378 if (e && !get_nat_flag(e, IS_CHECKPOINTED))
379 is_cp = false;
b873b798 380 up_read(&nm_i->nat_tree_lock);
2dcf51ab 381 return is_cp;
479f40c4
JK
382}
383
4d57b86d 384bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
b6fe5873
JK
385{
386 struct f2fs_nm_info *nm_i = NM_I(sbi);
387 struct nat_entry *e;
88bd02c9 388 bool need_update = true;
b6fe5873 389
b873b798 390 down_read(&nm_i->nat_tree_lock);
88bd02c9
JK
391 e = __lookup_nat_cache(nm_i, ino);
392 if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
393 (get_nat_flag(e, IS_CHECKPOINTED) ||
394 get_nat_flag(e, HAS_FSYNCED_INODE)))
395 need_update = false;
b873b798 396 up_read(&nm_i->nat_tree_lock);
88bd02c9 397 return need_update;
b6fe5873
JK
398}
399
12f9ef37 400/* must be locked by nat_tree_lock */
1515aef0 401static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
e05df3b1
JK
402 struct f2fs_nat_entry *ne)
403{
1515aef0 404 struct f2fs_nm_info *nm_i = NM_I(sbi);
12f9ef37 405 struct nat_entry *new, *e;
9be32d72 406
12f9ef37
YH
407 new = __alloc_nat_entry(nid, false);
408 if (!new)
409 return;
410
411 down_write(&nm_i->nat_tree_lock);
e05df3b1 412 e = __lookup_nat_cache(nm_i, nid);
12f9ef37
YH
413 if (!e)
414 e = __init_nat_entry(nm_i, new, ne, false);
415 else
0c0b471e
EB
416 f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) ||
417 nat_get_blkaddr(e) !=
418 le32_to_cpu(ne->block_addr) ||
1515aef0 419 nat_get_version(e) != ne->version);
12f9ef37
YH
420 up_write(&nm_i->nat_tree_lock);
421 if (e != new)
422 __free_nat_entry(new);
e05df3b1
JK
423}
424
425static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
479f40c4 426 block_t new_blkaddr, bool fsync_done)
e05df3b1
JK
427{
428 struct f2fs_nm_info *nm_i = NM_I(sbi);
429 struct nat_entry *e;
12f9ef37 430 struct nat_entry *new = __alloc_nat_entry(ni->nid, true);
9be32d72 431
b873b798 432 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK
433 e = __lookup_nat_cache(nm_i, ni->nid);
434 if (!e) {
12f9ef37 435 e = __init_nat_entry(nm_i, new, NULL, true);
5c27f4ee 436 copy_node_info(&e->ni, ni);
9850cf4a 437 f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR);
e05df3b1
JK
438 } else if (new_blkaddr == NEW_ADDR) {
439 /*
440 * when nid is reallocated,
441 * previous nat entry can be remained in nat cache.
442 * So, reinitialize it with new information.
443 */
5c27f4ee 444 copy_node_info(&e->ni, ni);
9850cf4a 445 f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR);
e05df3b1 446 }
12f9ef37
YH
447 /* let's free early to reduce memory consumption */
448 if (e != new)
449 __free_nat_entry(new);
e05df3b1 450
e05df3b1 451 /* sanity check */
9850cf4a
JK
452 f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr);
453 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR &&
e05df3b1 454 new_blkaddr == NULL_ADDR);
9850cf4a 455 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR &&
e05df3b1 456 new_blkaddr == NEW_ADDR);
e1da7872 457 f2fs_bug_on(sbi, is_valid_data_blkaddr(sbi, nat_get_blkaddr(e)) &&
e05df3b1
JK
458 new_blkaddr == NEW_ADDR);
459
e1c42045 460 /* increment version no as node is removed */
e05df3b1
JK
461 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
462 unsigned char version = nat_get_version(e);
463 nat_set_version(e, inc_node_version(version));
464 }
465
466 /* change address */
467 nat_set_blkaddr(e, new_blkaddr);
e1da7872 468 if (!is_valid_data_blkaddr(sbi, new_blkaddr))
88bd02c9 469 set_nat_flag(e, IS_CHECKPOINTED, false);
e05df3b1 470 __set_nat_cache_dirty(nm_i, e);
479f40c4
JK
471
472 /* update fsync_mark if its inode nat entry is still alive */
d5b692b7
CY
473 if (ni->nid != ni->ino)
474 e = __lookup_nat_cache(nm_i, ni->ino);
88bd02c9
JK
475 if (e) {
476 if (fsync_done && ni->nid == ni->ino)
477 set_nat_flag(e, HAS_FSYNCED_INODE, true);
478 set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
479 }
b873b798 480 up_write(&nm_i->nat_tree_lock);
e05df3b1
JK
481}
482
4d57b86d 483int f2fs_try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
e05df3b1
JK
484{
485 struct f2fs_nm_info *nm_i = NM_I(sbi);
1b38dc8e 486 int nr = nr_shrink;
e05df3b1 487
b873b798
JK
488 if (!down_write_trylock(&nm_i->nat_tree_lock))
489 return 0;
e05df3b1 490
22969158
CY
491 spin_lock(&nm_i->nat_list_lock);
492 while (nr_shrink) {
e05df3b1 493 struct nat_entry *ne;
22969158
CY
494
495 if (list_empty(&nm_i->nat_entries))
496 break;
497
e05df3b1
JK
498 ne = list_first_entry(&nm_i->nat_entries,
499 struct nat_entry, list);
22969158
CY
500 list_del(&ne->list);
501 spin_unlock(&nm_i->nat_list_lock);
502
e05df3b1
JK
503 __del_from_nat_cache(nm_i, ne);
504 nr_shrink--;
22969158
CY
505
506 spin_lock(&nm_i->nat_list_lock);
e05df3b1 507 }
22969158
CY
508 spin_unlock(&nm_i->nat_list_lock);
509
b873b798 510 up_write(&nm_i->nat_tree_lock);
1b38dc8e 511 return nr - nr_shrink;
e05df3b1
JK
512}
513
0a8165d7 514/*
e1c42045 515 * This function always returns success
e05df3b1 516 */
7735730d 517int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
4d57b86d 518 struct node_info *ni)
e05df3b1
JK
519{
520 struct f2fs_nm_info *nm_i = NM_I(sbi);
521 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 522 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK
523 nid_t start_nid = START_NID(nid);
524 struct f2fs_nat_block *nat_blk;
525 struct page *page = NULL;
526 struct f2fs_nat_entry ne;
527 struct nat_entry *e;
66a82d1f 528 pgoff_t index;
e05df3b1
JK
529 int i;
530
531 ni->nid = nid;
532
533 /* Check nat cache */
b873b798 534 down_read(&nm_i->nat_tree_lock);
e05df3b1
JK
535 e = __lookup_nat_cache(nm_i, nid);
536 if (e) {
537 ni->ino = nat_get_ino(e);
538 ni->blk_addr = nat_get_blkaddr(e);
539 ni->version = nat_get_version(e);
b873b798 540 up_read(&nm_i->nat_tree_lock);
7735730d 541 return 0;
1515aef0 542 }
e05df3b1 543
3547ea96
JK
544 memset(&ne, 0, sizeof(struct f2fs_nat_entry));
545
e05df3b1 546 /* Check current segment summary */
b7ad7512 547 down_read(&curseg->journal_rwsem);
4d57b86d 548 i = f2fs_lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0);
e05df3b1 549 if (i >= 0) {
dfc08a12 550 ne = nat_in_journal(journal, i);
e05df3b1
JK
551 node_info_from_raw_nat(ni, &ne);
552 }
b7ad7512 553 up_read(&curseg->journal_rwsem);
66a82d1f
YH
554 if (i >= 0) {
555 up_read(&nm_i->nat_tree_lock);
e05df3b1 556 goto cache;
66a82d1f 557 }
e05df3b1
JK
558
559 /* Fill node_info from nat page */
66a82d1f
YH
560 index = current_nat_addr(sbi, nid);
561 up_read(&nm_i->nat_tree_lock);
562
4d57b86d 563 page = f2fs_get_meta_page(sbi, index);
7735730d
CY
564 if (IS_ERR(page))
565 return PTR_ERR(page);
566
e05df3b1
JK
567 nat_blk = (struct f2fs_nat_block *)page_address(page);
568 ne = nat_blk->entries[nid - start_nid];
569 node_info_from_raw_nat(ni, &ne);
570 f2fs_put_page(page, 1);
571cache:
572 /* cache nat entry */
1515aef0 573 cache_nat_entry(sbi, nid, &ne);
7735730d 574 return 0;
e05df3b1
JK
575}
576
79344efb
JK
577/*
578 * readahead MAX_RA_NODE number of node pages.
579 */
4d57b86d 580static void f2fs_ra_node_pages(struct page *parent, int start, int n)
79344efb
JK
581{
582 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
583 struct blk_plug plug;
584 int i, end;
585 nid_t nid;
586
587 blk_start_plug(&plug);
588
589 /* Then, try readahead for siblings of the desired node */
590 end = start + n;
591 end = min(end, NIDS_PER_BLOCK);
592 for (i = start; i < end; i++) {
593 nid = get_nid(parent, i, false);
4d57b86d 594 f2fs_ra_node_page(sbi, nid);
79344efb
JK
595 }
596
597 blk_finish_plug(&plug);
598}
599
4d57b86d 600pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs)
3cf45747
CY
601{
602 const long direct_index = ADDRS_PER_INODE(dn->inode);
603 const long direct_blks = ADDRS_PER_BLOCK;
604 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
605 unsigned int skipped_unit = ADDRS_PER_BLOCK;
606 int cur_level = dn->cur_level;
607 int max_level = dn->max_level;
608 pgoff_t base = 0;
609
610 if (!dn->max_level)
611 return pgofs + 1;
612
613 while (max_level-- > cur_level)
614 skipped_unit *= NIDS_PER_BLOCK;
615
616 switch (dn->max_level) {
617 case 3:
618 base += 2 * indirect_blks;
619 case 2:
620 base += 2 * direct_blks;
621 case 1:
622 base += direct_index;
623 break;
624 default:
625 f2fs_bug_on(F2FS_I_SB(dn->inode), 1);
626 }
627
628 return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base;
629}
630
0a8165d7 631/*
e05df3b1
JK
632 * The maximum depth is four.
633 * Offset[0] will have raw inode offset.
634 */
81ca7350 635static int get_node_path(struct inode *inode, long block,
de93653f 636 int offset[4], unsigned int noffset[4])
e05df3b1 637{
81ca7350 638 const long direct_index = ADDRS_PER_INODE(inode);
e05df3b1
JK
639 const long direct_blks = ADDRS_PER_BLOCK;
640 const long dptrs_per_blk = NIDS_PER_BLOCK;
641 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK;
642 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
643 int n = 0;
644 int level = 0;
645
646 noffset[0] = 0;
647
648 if (block < direct_index) {
25c0a6e5 649 offset[n] = block;
e05df3b1
JK
650 goto got;
651 }
652 block -= direct_index;
653 if (block < direct_blks) {
654 offset[n++] = NODE_DIR1_BLOCK;
655 noffset[n] = 1;
25c0a6e5 656 offset[n] = block;
e05df3b1
JK
657 level = 1;
658 goto got;
659 }
660 block -= direct_blks;
661 if (block < direct_blks) {
662 offset[n++] = NODE_DIR2_BLOCK;
663 noffset[n] = 2;
25c0a6e5 664 offset[n] = block;
e05df3b1
JK
665 level = 1;
666 goto got;
667 }
668 block -= direct_blks;
669 if (block < indirect_blks) {
670 offset[n++] = NODE_IND1_BLOCK;
671 noffset[n] = 3;
672 offset[n++] = block / direct_blks;
673 noffset[n] = 4 + offset[n - 1];
25c0a6e5 674 offset[n] = block % direct_blks;
e05df3b1
JK
675 level = 2;
676 goto got;
677 }
678 block -= indirect_blks;
679 if (block < indirect_blks) {
680 offset[n++] = NODE_IND2_BLOCK;
681 noffset[n] = 4 + dptrs_per_blk;
682 offset[n++] = block / direct_blks;
683 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
25c0a6e5 684 offset[n] = block % direct_blks;
e05df3b1
JK
685 level = 2;
686 goto got;
687 }
688 block -= indirect_blks;
689 if (block < dindirect_blks) {
690 offset[n++] = NODE_DIND_BLOCK;
691 noffset[n] = 5 + (dptrs_per_blk * 2);
692 offset[n++] = block / indirect_blks;
693 noffset[n] = 6 + (dptrs_per_blk * 2) +
694 offset[n - 1] * (dptrs_per_blk + 1);
695 offset[n++] = (block / direct_blks) % dptrs_per_blk;
696 noffset[n] = 7 + (dptrs_per_blk * 2) +
697 offset[n - 2] * (dptrs_per_blk + 1) +
698 offset[n - 1];
25c0a6e5 699 offset[n] = block % direct_blks;
e05df3b1
JK
700 level = 3;
701 goto got;
702 } else {
adb6dc19 703 return -E2BIG;
e05df3b1
JK
704 }
705got:
706 return level;
707}
708
709/*
710 * Caller should call f2fs_put_dnode(dn).
4f4124d0
CY
711 * Also, it should grab and release a rwsem by calling f2fs_lock_op() and
712 * f2fs_unlock_op() only if ro is not set RDONLY_NODE.
39936837 713 * In the case of RDONLY_NODE, we don't need to care about mutex.
e05df3b1 714 */
4d57b86d 715int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
e05df3b1 716{
4081363f 717 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 718 struct page *npage[4];
f1a3b98e 719 struct page *parent = NULL;
e05df3b1
JK
720 int offset[4];
721 unsigned int noffset[4];
722 nid_t nids[4];
3cf45747 723 int level, i = 0;
e05df3b1
JK
724 int err = 0;
725
81ca7350 726 level = get_node_path(dn->inode, index, offset, noffset);
adb6dc19
JK
727 if (level < 0)
728 return level;
e05df3b1
JK
729
730 nids[0] = dn->inode->i_ino;
1646cfac 731 npage[0] = dn->inode_page;
e05df3b1 732
1646cfac 733 if (!npage[0]) {
4d57b86d 734 npage[0] = f2fs_get_node_page(sbi, nids[0]);
1646cfac
JK
735 if (IS_ERR(npage[0]))
736 return PTR_ERR(npage[0]);
737 }
f1a3b98e
JK
738
739 /* if inline_data is set, should not report any block indices */
740 if (f2fs_has_inline_data(dn->inode) && index) {
76629165 741 err = -ENOENT;
f1a3b98e
JK
742 f2fs_put_page(npage[0], 1);
743 goto release_out;
744 }
745
e05df3b1 746 parent = npage[0];
52c2db3f
CL
747 if (level != 0)
748 nids[1] = get_nid(parent, offset[0], true);
e05df3b1
JK
749 dn->inode_page = npage[0];
750 dn->inode_page_locked = true;
751
752 /* get indirect or direct nodes */
753 for (i = 1; i <= level; i++) {
754 bool done = false;
755
266e97a8 756 if (!nids[i] && mode == ALLOC_NODE) {
e05df3b1 757 /* alloc new node */
4d57b86d 758 if (!f2fs_alloc_nid(sbi, &(nids[i]))) {
e05df3b1
JK
759 err = -ENOSPC;
760 goto release_pages;
761 }
762
763 dn->nid = nids[i];
4d57b86d 764 npage[i] = f2fs_new_node_page(dn, noffset[i]);
e05df3b1 765 if (IS_ERR(npage[i])) {
4d57b86d 766 f2fs_alloc_nid_failed(sbi, nids[i]);
e05df3b1
JK
767 err = PTR_ERR(npage[i]);
768 goto release_pages;
769 }
770
771 set_nid(parent, offset[i - 1], nids[i], i == 1);
4d57b86d 772 f2fs_alloc_nid_done(sbi, nids[i]);
e05df3b1 773 done = true;
266e97a8 774 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
4d57b86d 775 npage[i] = f2fs_get_node_page_ra(parent, offset[i - 1]);
e05df3b1
JK
776 if (IS_ERR(npage[i])) {
777 err = PTR_ERR(npage[i]);
778 goto release_pages;
779 }
780 done = true;
781 }
782 if (i == 1) {
783 dn->inode_page_locked = false;
784 unlock_page(parent);
785 } else {
786 f2fs_put_page(parent, 1);
787 }
788
789 if (!done) {
4d57b86d 790 npage[i] = f2fs_get_node_page(sbi, nids[i]);
e05df3b1
JK
791 if (IS_ERR(npage[i])) {
792 err = PTR_ERR(npage[i]);
793 f2fs_put_page(npage[0], 0);
794 goto release_out;
795 }
796 }
797 if (i < level) {
798 parent = npage[i];
799 nids[i + 1] = get_nid(parent, offset[i], false);
800 }
801 }
802 dn->nid = nids[level];
803 dn->ofs_in_node = offset[level];
804 dn->node_page = npage[level];
7a2af766
CY
805 dn->data_blkaddr = datablock_addr(dn->inode,
806 dn->node_page, dn->ofs_in_node);
e05df3b1
JK
807 return 0;
808
809release_pages:
810 f2fs_put_page(parent, 1);
811 if (i > 1)
812 f2fs_put_page(npage[0], 0);
813release_out:
814 dn->inode_page = NULL;
815 dn->node_page = NULL;
3cf45747
CY
816 if (err == -ENOENT) {
817 dn->cur_level = i;
818 dn->max_level = level;
0a2aa8fb 819 dn->ofs_in_node = offset[level];
3cf45747 820 }
e05df3b1
JK
821 return err;
822}
823
7735730d 824static int truncate_node(struct dnode_of_data *dn)
e05df3b1 825{
4081363f 826 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 827 struct node_info ni;
7735730d 828 int err;
0ea295dd 829 pgoff_t index;
e05df3b1 830
7735730d
CY
831 err = f2fs_get_node_info(sbi, dn->nid, &ni);
832 if (err)
833 return err;
e05df3b1 834
e05df3b1 835 /* Deallocate node address */
4d57b86d 836 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 837 dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino);
479f40c4 838 set_node_addr(sbi, &ni, NULL_ADDR, false);
e05df3b1
JK
839
840 if (dn->nid == dn->inode->i_ino) {
4d57b86d 841 f2fs_remove_orphan_inode(sbi, dn->nid);
e05df3b1 842 dec_valid_inode_count(sbi);
0f18b462 843 f2fs_inode_synced(dn->inode);
e05df3b1 844 }
000519f2 845
e05df3b1 846 clear_node_page_dirty(dn->node_page);
caf0047e 847 set_sbi_flag(sbi, SBI_IS_DIRTY);
e05df3b1 848
0ea295dd 849 index = dn->node_page->index;
e05df3b1 850 f2fs_put_page(dn->node_page, 1);
bf39c00a
JK
851
852 invalidate_mapping_pages(NODE_MAPPING(sbi),
0ea295dd 853 index, index);
bf39c00a 854
e05df3b1 855 dn->node_page = NULL;
51dd6249 856 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
7735730d
CY
857
858 return 0;
e05df3b1
JK
859}
860
861static int truncate_dnode(struct dnode_of_data *dn)
862{
e05df3b1 863 struct page *page;
7735730d 864 int err;
e05df3b1
JK
865
866 if (dn->nid == 0)
867 return 1;
868
869 /* get direct node */
4d57b86d 870 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
e05df3b1
JK
871 if (IS_ERR(page) && PTR_ERR(page) == -ENOENT)
872 return 1;
873 else if (IS_ERR(page))
874 return PTR_ERR(page);
875
876 /* Make dnode_of_data for parameter */
877 dn->node_page = page;
878 dn->ofs_in_node = 0;
4d57b86d 879 f2fs_truncate_data_blocks(dn);
7735730d
CY
880 err = truncate_node(dn);
881 if (err)
882 return err;
883
e05df3b1
JK
884 return 1;
885}
886
887static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
888 int ofs, int depth)
889{
e05df3b1
JK
890 struct dnode_of_data rdn = *dn;
891 struct page *page;
892 struct f2fs_node *rn;
893 nid_t child_nid;
894 unsigned int child_nofs;
895 int freed = 0;
896 int i, ret;
897
898 if (dn->nid == 0)
899 return NIDS_PER_BLOCK + 1;
900
51dd6249
NJ
901 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
902
4d57b86d 903 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
51dd6249
NJ
904 if (IS_ERR(page)) {
905 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
e05df3b1 906 return PTR_ERR(page);
51dd6249 907 }
e05df3b1 908
4d57b86d 909 f2fs_ra_node_pages(page, ofs, NIDS_PER_BLOCK);
79344efb 910
45590710 911 rn = F2FS_NODE(page);
e05df3b1
JK
912 if (depth < 3) {
913 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
914 child_nid = le32_to_cpu(rn->in.nid[i]);
915 if (child_nid == 0)
916 continue;
917 rdn.nid = child_nid;
918 ret = truncate_dnode(&rdn);
919 if (ret < 0)
920 goto out_err;
12719ae1
JK
921 if (set_nid(page, i, 0, false))
922 dn->node_changed = true;
e05df3b1
JK
923 }
924 } else {
925 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
926 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
927 child_nid = le32_to_cpu(rn->in.nid[i]);
928 if (child_nid == 0) {
929 child_nofs += NIDS_PER_BLOCK + 1;
930 continue;
931 }
932 rdn.nid = child_nid;
933 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
934 if (ret == (NIDS_PER_BLOCK + 1)) {
12719ae1
JK
935 if (set_nid(page, i, 0, false))
936 dn->node_changed = true;
e05df3b1
JK
937 child_nofs += ret;
938 } else if (ret < 0 && ret != -ENOENT) {
939 goto out_err;
940 }
941 }
942 freed = child_nofs;
943 }
944
945 if (!ofs) {
946 /* remove current indirect node */
947 dn->node_page = page;
7735730d
CY
948 ret = truncate_node(dn);
949 if (ret)
950 goto out_err;
e05df3b1
JK
951 freed++;
952 } else {
953 f2fs_put_page(page, 1);
954 }
51dd6249 955 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
e05df3b1
JK
956 return freed;
957
958out_err:
959 f2fs_put_page(page, 1);
51dd6249 960 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
e05df3b1
JK
961 return ret;
962}
963
964static int truncate_partial_nodes(struct dnode_of_data *dn,
965 struct f2fs_inode *ri, int *offset, int depth)
966{
e05df3b1
JK
967 struct page *pages[2];
968 nid_t nid[3];
969 nid_t child_nid;
970 int err = 0;
971 int i;
972 int idx = depth - 2;
973
974 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
975 if (!nid[0])
976 return 0;
977
978 /* get indirect nodes in the path */
a225dca3 979 for (i = 0; i < idx + 1; i++) {
e1c42045 980 /* reference count'll be increased */
4d57b86d 981 pages[i] = f2fs_get_node_page(F2FS_I_SB(dn->inode), nid[i]);
e05df3b1 982 if (IS_ERR(pages[i])) {
e05df3b1 983 err = PTR_ERR(pages[i]);
a225dca3 984 idx = i - 1;
e05df3b1
JK
985 goto fail;
986 }
987 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
988 }
989
4d57b86d 990 f2fs_ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK);
79344efb 991
e05df3b1 992 /* free direct nodes linked to a partial indirect node */
a225dca3 993 for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) {
e05df3b1
JK
994 child_nid = get_nid(pages[idx], i, false);
995 if (!child_nid)
996 continue;
997 dn->nid = child_nid;
998 err = truncate_dnode(dn);
999 if (err < 0)
1000 goto fail;
12719ae1
JK
1001 if (set_nid(pages[idx], i, 0, false))
1002 dn->node_changed = true;
e05df3b1
JK
1003 }
1004
a225dca3 1005 if (offset[idx + 1] == 0) {
e05df3b1
JK
1006 dn->node_page = pages[idx];
1007 dn->nid = nid[idx];
7735730d
CY
1008 err = truncate_node(dn);
1009 if (err)
1010 goto fail;
e05df3b1
JK
1011 } else {
1012 f2fs_put_page(pages[idx], 1);
1013 }
1014 offset[idx]++;
a225dca3 1015 offset[idx + 1] = 0;
1016 idx--;
e05df3b1 1017fail:
a225dca3 1018 for (i = idx; i >= 0; i--)
e05df3b1 1019 f2fs_put_page(pages[i], 1);
51dd6249
NJ
1020
1021 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
1022
e05df3b1
JK
1023 return err;
1024}
1025
0a8165d7 1026/*
e05df3b1
JK
1027 * All the block addresses of data and nodes should be nullified.
1028 */
4d57b86d 1029int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from)
e05df3b1 1030{
4081363f 1031 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e05df3b1
JK
1032 int err = 0, cont = 1;
1033 int level, offset[4], noffset[4];
7dd690c8 1034 unsigned int nofs = 0;
58bfaf44 1035 struct f2fs_inode *ri;
e05df3b1
JK
1036 struct dnode_of_data dn;
1037 struct page *page;
1038
51dd6249
NJ
1039 trace_f2fs_truncate_inode_blocks_enter(inode, from);
1040
81ca7350 1041 level = get_node_path(inode, from, offset, noffset);
adb6dc19
JK
1042 if (level < 0)
1043 return level;
ff373558 1044
4d57b86d 1045 page = f2fs_get_node_page(sbi, inode->i_ino);
51dd6249
NJ
1046 if (IS_ERR(page)) {
1047 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
e05df3b1 1048 return PTR_ERR(page);
51dd6249 1049 }
e05df3b1
JK
1050
1051 set_new_dnode(&dn, inode, page, NULL, 0);
1052 unlock_page(page);
1053
58bfaf44 1054 ri = F2FS_INODE(page);
e05df3b1
JK
1055 switch (level) {
1056 case 0:
1057 case 1:
1058 nofs = noffset[1];
1059 break;
1060 case 2:
1061 nofs = noffset[1];
1062 if (!offset[level - 1])
1063 goto skip_partial;
58bfaf44 1064 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK
1065 if (err < 0 && err != -ENOENT)
1066 goto fail;
1067 nofs += 1 + NIDS_PER_BLOCK;
1068 break;
1069 case 3:
1070 nofs = 5 + 2 * NIDS_PER_BLOCK;
1071 if (!offset[level - 1])
1072 goto skip_partial;
58bfaf44 1073 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK
1074 if (err < 0 && err != -ENOENT)
1075 goto fail;
1076 break;
1077 default:
1078 BUG();
1079 }
1080
1081skip_partial:
1082 while (cont) {
58bfaf44 1083 dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
e05df3b1
JK
1084 switch (offset[0]) {
1085 case NODE_DIR1_BLOCK:
1086 case NODE_DIR2_BLOCK:
1087 err = truncate_dnode(&dn);
1088 break;
1089
1090 case NODE_IND1_BLOCK:
1091 case NODE_IND2_BLOCK:
1092 err = truncate_nodes(&dn, nofs, offset[1], 2);
1093 break;
1094
1095 case NODE_DIND_BLOCK:
1096 err = truncate_nodes(&dn, nofs, offset[1], 3);
1097 cont = 0;
1098 break;
1099
1100 default:
1101 BUG();
1102 }
1103 if (err < 0 && err != -ENOENT)
1104 goto fail;
1105 if (offset[1] == 0 &&
58bfaf44 1106 ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) {
e05df3b1 1107 lock_page(page);
ff373558 1108 BUG_ON(page->mapping != NODE_MAPPING(sbi));
fec1d657 1109 f2fs_wait_on_page_writeback(page, NODE, true);
58bfaf44 1110 ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
e05df3b1
JK
1111 set_page_dirty(page);
1112 unlock_page(page);
1113 }
1114 offset[1] = 0;
1115 offset[0]++;
1116 nofs += err;
1117 }
1118fail:
1119 f2fs_put_page(page, 0);
51dd6249 1120 trace_f2fs_truncate_inode_blocks_exit(inode, err);
e05df3b1
JK
1121 return err > 0 ? 0 : err;
1122}
1123
9c77f754 1124/* caller must lock inode page */
4d57b86d 1125int f2fs_truncate_xattr_node(struct inode *inode)
4f16fb0f 1126{
4081363f 1127 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4f16fb0f
JK
1128 nid_t nid = F2FS_I(inode)->i_xattr_nid;
1129 struct dnode_of_data dn;
1130 struct page *npage;
7735730d 1131 int err;
4f16fb0f
JK
1132
1133 if (!nid)
1134 return 0;
1135
4d57b86d 1136 npage = f2fs_get_node_page(sbi, nid);
4f16fb0f
JK
1137 if (IS_ERR(npage))
1138 return PTR_ERR(npage);
1139
7735730d
CY
1140 set_new_dnode(&dn, inode, NULL, npage, nid);
1141 err = truncate_node(&dn);
1142 if (err) {
1143 f2fs_put_page(npage, 1);
1144 return err;
1145 }
1146
205b9822 1147 f2fs_i_xnid_write(inode, 0);
65985d93 1148
4f16fb0f
JK
1149 return 0;
1150}
1151
39936837 1152/*
4f4124d0
CY
1153 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
1154 * f2fs_unlock_op().
39936837 1155 */
4d57b86d 1156int f2fs_remove_inode_page(struct inode *inode)
e05df3b1 1157{
e05df3b1 1158 struct dnode_of_data dn;
13ec7297 1159 int err;
e05df3b1 1160
c2e69583 1161 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
4d57b86d 1162 err = f2fs_get_dnode_of_data(&dn, 0, LOOKUP_NODE);
13ec7297
CY
1163 if (err)
1164 return err;
e05df3b1 1165
4d57b86d 1166 err = f2fs_truncate_xattr_node(inode);
13ec7297 1167 if (err) {
c2e69583 1168 f2fs_put_dnode(&dn);
13ec7297 1169 return err;
e05df3b1 1170 }
c2e69583
JK
1171
1172 /* remove potential inline_data blocks */
1173 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1174 S_ISLNK(inode->i_mode))
4d57b86d 1175 f2fs_truncate_data_blocks_range(&dn, 1);
c2e69583 1176
e1c42045 1177 /* 0 is possible, after f2fs_new_inode() has failed */
8d714f8a
JK
1178 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
1179 f2fs_put_dnode(&dn);
1180 return -EIO;
1181 }
9850cf4a 1182 f2fs_bug_on(F2FS_I_SB(inode),
0eb0adad 1183 inode->i_blocks != 0 && inode->i_blocks != 8);
c2e69583
JK
1184
1185 /* will put inode & node pages */
7735730d
CY
1186 err = truncate_node(&dn);
1187 if (err) {
1188 f2fs_put_dnode(&dn);
1189 return err;
1190 }
13ec7297 1191 return 0;
e05df3b1
JK
1192}
1193
4d57b86d 1194struct page *f2fs_new_inode_page(struct inode *inode)
e05df3b1 1195{
e05df3b1
JK
1196 struct dnode_of_data dn;
1197
1198 /* allocate inode page for new inode */
1199 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
44a83ff6
JK
1200
1201 /* caller should f2fs_put_page(page, 1); */
4d57b86d 1202 return f2fs_new_node_page(&dn, 0);
e05df3b1
JK
1203}
1204
4d57b86d 1205struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs)
e05df3b1 1206{
4081363f 1207 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
25cc5d3b 1208 struct node_info new_ni;
e05df3b1
JK
1209 struct page *page;
1210 int err;
1211
91942321 1212 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
e05df3b1
JK
1213 return ERR_PTR(-EPERM);
1214
300e129c 1215 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false);
e05df3b1
JK
1216 if (!page)
1217 return ERR_PTR(-ENOMEM);
1218
0abd675e 1219 if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs))))
9c02740c 1220 goto fail;
0abd675e 1221
25cc5d3b 1222#ifdef CONFIG_F2FS_CHECK_FS
7735730d
CY
1223 err = f2fs_get_node_info(sbi, dn->nid, &new_ni);
1224 if (err) {
1225 dec_valid_node_count(sbi, dn->inode, !ofs);
1226 goto fail;
1227 }
25cc5d3b
JK
1228 f2fs_bug_on(sbi, new_ni.blk_addr != NULL_ADDR);
1229#endif
1230 new_ni.nid = dn->nid;
e05df3b1 1231 new_ni.ino = dn->inode->i_ino;
25cc5d3b
JK
1232 new_ni.blk_addr = NULL_ADDR;
1233 new_ni.flag = 0;
1234 new_ni.version = 0;
479f40c4 1235 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
9c02740c 1236
fec1d657 1237 f2fs_wait_on_page_writeback(page, NODE, true);
9c02740c 1238 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
c5667575 1239 set_cold_node(page, S_ISDIR(dn->inode->i_mode));
237c0790
JK
1240 if (!PageUptodate(page))
1241 SetPageUptodate(page);
12719ae1
JK
1242 if (set_page_dirty(page))
1243 dn->node_changed = true;
e05df3b1 1244
4bc8e9bc 1245 if (f2fs_has_xattr_block(ofs))
205b9822 1246 f2fs_i_xnid_write(dn->inode, dn->nid);
479bd73a 1247
e05df3b1
JK
1248 if (ofs == 0)
1249 inc_valid_inode_count(sbi);
e05df3b1
JK
1250 return page;
1251
1252fail:
71e9fec5 1253 clear_node_page_dirty(page);
e05df3b1
JK
1254 f2fs_put_page(page, 1);
1255 return ERR_PTR(err);
1256}
1257
56ae674c
JK
1258/*
1259 * Caller should do after getting the following values.
1260 * 0: f2fs_put_page(page, 0)
86531d6b 1261 * LOCKED_PAGE or error: f2fs_put_page(page, 1)
56ae674c 1262 */
04d328de 1263static int read_node_page(struct page *page, int op_flags)
e05df3b1 1264{
4081363f 1265 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
e05df3b1 1266 struct node_info ni;
cf04e8eb 1267 struct f2fs_io_info fio = {
05ca3632 1268 .sbi = sbi,
cf04e8eb 1269 .type = NODE,
04d328de
MC
1270 .op = REQ_OP_READ,
1271 .op_flags = op_flags,
05ca3632 1272 .page = page,
4375a336 1273 .encrypted_page = NULL,
cf04e8eb 1274 };
7735730d 1275 int err;
e05df3b1 1276
54c55c4e
WG
1277 if (PageUptodate(page)) {
1278#ifdef CONFIG_F2FS_CHECK_FS
1279 f2fs_bug_on(sbi, !f2fs_inode_chksum_verify(sbi, page));
1280#endif
3bdad3c7 1281 return LOCKED_PAGE;
54c55c4e 1282 }
3bdad3c7 1283
7735730d
CY
1284 err = f2fs_get_node_info(sbi, page->index, &ni);
1285 if (err)
1286 return err;
e05df3b1 1287
83a3bfdb
JK
1288 if (unlikely(ni.blk_addr == NULL_ADDR) ||
1289 is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) {
2bca1e23 1290 ClearPageUptodate(page);
e05df3b1 1291 return -ENOENT;
393ff91f
JK
1292 }
1293
7a9d7548 1294 fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr;
05ca3632 1295 return f2fs_submit_page_bio(&fio);
e05df3b1
JK
1296}
1297
0a8165d7 1298/*
e05df3b1
JK
1299 * Readahead a node page
1300 */
4d57b86d 1301void f2fs_ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 1302{
e05df3b1 1303 struct page *apage;
56ae674c 1304 int err;
e05df3b1 1305
e8458725
CY
1306 if (!nid)
1307 return;
4d57b86d 1308 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1309 return;
e8458725 1310
5ec2d99d 1311 apage = xa_load(&NODE_MAPPING(sbi)->i_pages, nid);
999270de 1312 if (apage)
393ff91f 1313 return;
e05df3b1 1314
300e129c 1315 apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK
1316 if (!apage)
1317 return;
1318
70246286 1319 err = read_node_page(apage, REQ_RAHEAD);
86531d6b 1320 f2fs_put_page(apage, err ? 1 : 0);
e05df3b1
JK
1321}
1322
17a0ee55 1323static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid,
0e022ea8 1324 struct page *parent, int start)
e05df3b1 1325{
e05df3b1 1326 struct page *page;
0e022ea8 1327 int err;
e05df3b1 1328
e05df3b1
JK
1329 if (!nid)
1330 return ERR_PTR(-ENOENT);
4d57b86d 1331 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1332 return ERR_PTR(-EINVAL);
afcb7ca0 1333repeat:
300e129c 1334 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK
1335 if (!page)
1336 return ERR_PTR(-ENOMEM);
1337
70fd7614 1338 err = read_node_page(page, 0);
86531d6b
JK
1339 if (err < 0) {
1340 f2fs_put_page(page, 1);
e05df3b1 1341 return ERR_PTR(err);
86531d6b 1342 } else if (err == LOCKED_PAGE) {
1f258ec1 1343 err = 0;
56ae674c 1344 goto page_hit;
86531d6b 1345 }
e05df3b1 1346
0e022ea8 1347 if (parent)
4d57b86d 1348 f2fs_ra_node_pages(parent, start + 1, MAX_RA_NODE);
c718379b 1349
e05df3b1 1350 lock_page(page);
0e022ea8 1351
4ef51a8f 1352 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
afcb7ca0
JK
1353 f2fs_put_page(page, 1);
1354 goto repeat;
1355 }
1563ac75 1356
1f258ec1
CY
1357 if (unlikely(!PageUptodate(page))) {
1358 err = -EIO;
1563ac75 1359 goto out_err;
1f258ec1 1360 }
704956ec
CY
1361
1362 if (!f2fs_inode_chksum_verify(sbi, page)) {
1363 err = -EBADMSG;
1364 goto out_err;
1365 }
e0f56cb4 1366page_hit:
0c9df7fb 1367 if(unlikely(nid != nid_of_node(page))) {
1f258ec1
CY
1368 f2fs_msg(sbi->sb, KERN_WARNING, "inconsistent node block, "
1369 "nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]",
1370 nid, nid_of_node(page), ino_of_node(page),
1371 ofs_of_node(page), cpver_of_node(page),
1372 next_blkaddr_of_node(page));
1f258ec1 1373 err = -EINVAL;
0c9df7fb 1374out_err:
ee605234 1375 ClearPageUptodate(page);
0c9df7fb 1376 f2fs_put_page(page, 1);
1f258ec1 1377 return ERR_PTR(err);
0c9df7fb 1378 }
e05df3b1
JK
1379 return page;
1380}
1381
4d57b86d 1382struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
0e022ea8
CY
1383{
1384 return __get_node_page(sbi, nid, NULL, 0);
1385}
1386
4d57b86d 1387struct page *f2fs_get_node_page_ra(struct page *parent, int start)
0e022ea8
CY
1388{
1389 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
1390 nid_t nid = get_nid(parent, start, false);
1391
1392 return __get_node_page(sbi, nid, parent, start);
1393}
1394
2049d4fc
JK
1395static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino)
1396{
1397 struct inode *inode;
1398 struct page *page;
0f3311a8 1399 int ret;
2049d4fc
JK
1400
1401 /* should flush inline_data before evict_inode */
1402 inode = ilookup(sbi->sb, ino);
1403 if (!inode)
1404 return;
1405
01eccef7
CY
1406 page = f2fs_pagecache_get_page(inode->i_mapping, 0,
1407 FGP_LOCK|FGP_NOWAIT, 0);
2049d4fc
JK
1408 if (!page)
1409 goto iput_out;
1410
1411 if (!PageUptodate(page))
1412 goto page_out;
1413
1414 if (!PageDirty(page))
1415 goto page_out;
1416
1417 if (!clear_page_dirty_for_io(page))
1418 goto page_out;
1419
0f3311a8
CY
1420 ret = f2fs_write_inline_data(inode, page);
1421 inode_dec_dirty_pages(inode);
4d57b86d 1422 f2fs_remove_dirty_inode(inode);
0f3311a8 1423 if (ret)
2049d4fc
JK
1424 set_page_dirty(page);
1425page_out:
4a6de50d 1426 f2fs_put_page(page, 1);
2049d4fc
JK
1427iput_out:
1428 iput(inode);
1429}
1430
608514de
JK
1431static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
1432{
028a63a6 1433 pgoff_t index;
608514de
JK
1434 struct pagevec pvec;
1435 struct page *last_page = NULL;
028a63a6 1436 int nr_pages;
608514de 1437
86679820 1438 pagevec_init(&pvec);
608514de 1439 index = 0;
028a63a6
JK
1440
1441 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1442 PAGECACHE_TAG_DIRTY))) {
028a63a6 1443 int i;
608514de
JK
1444
1445 for (i = 0; i < nr_pages; i++) {
1446 struct page *page = pvec.pages[i];
1447
1448 if (unlikely(f2fs_cp_error(sbi))) {
1449 f2fs_put_page(last_page, 0);
1450 pagevec_release(&pvec);
1451 return ERR_PTR(-EIO);
1452 }
1453
1454 if (!IS_DNODE(page) || !is_cold_node(page))
1455 continue;
1456 if (ino_of_node(page) != ino)
1457 continue;
1458
1459 lock_page(page);
1460
1461 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1462continue_unlock:
1463 unlock_page(page);
1464 continue;
1465 }
1466 if (ino_of_node(page) != ino)
1467 goto continue_unlock;
1468
1469 if (!PageDirty(page)) {
1470 /* someone wrote it for us */
1471 goto continue_unlock;
1472 }
1473
1474 if (last_page)
1475 f2fs_put_page(last_page, 0);
1476
1477 get_page(page);
1478 last_page = page;
1479 unlock_page(page);
1480 }
1481 pagevec_release(&pvec);
1482 cond_resched();
1483 }
1484 return last_page;
1485}
1486
d68f735b 1487static int __write_node_page(struct page *page, bool atomic, bool *submitted,
b0af6d49 1488 struct writeback_control *wbc, bool do_balance,
50fa53ec 1489 enum iostat_type io_type, unsigned int *seq_id)
faa24895
JK
1490{
1491 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1492 nid_t nid;
1493 struct node_info ni;
1494 struct f2fs_io_info fio = {
1495 .sbi = sbi,
39d787be 1496 .ino = ino_of_node(page),
faa24895
JK
1497 .type = NODE,
1498 .op = REQ_OP_WRITE,
1499 .op_flags = wbc_to_write_flags(wbc),
1500 .page = page,
1501 .encrypted_page = NULL,
d68f735b 1502 .submitted = false,
b0af6d49 1503 .io_type = io_type,
578c6478 1504 .io_wbc = wbc,
faa24895 1505 };
50fa53ec 1506 unsigned int seq;
faa24895
JK
1507
1508 trace_f2fs_writepage(page, NODE);
1509
868de613
JK
1510 if (unlikely(f2fs_cp_error(sbi)))
1511 goto redirty_out;
db198ae0 1512
faa24895
JK
1513 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1514 goto redirty_out;
faa24895 1515
fd8c8caf
CY
1516 if (wbc->sync_mode == WB_SYNC_NONE &&
1517 IS_DNODE(page) && is_cold_node(page))
1518 goto redirty_out;
1519
faa24895
JK
1520 /* get old block addr of this node page */
1521 nid = nid_of_node(page);
1522 f2fs_bug_on(sbi, page->index != nid);
1523
7735730d
CY
1524 if (f2fs_get_node_info(sbi, nid, &ni))
1525 goto redirty_out;
1526
faa24895
JK
1527 if (wbc->for_reclaim) {
1528 if (!down_read_trylock(&sbi->node_write))
1529 goto redirty_out;
1530 } else {
1531 down_read(&sbi->node_write);
1532 }
1533
faa24895
JK
1534 /* This page is already truncated */
1535 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1536 ClearPageUptodate(page);
1537 dec_page_count(sbi, F2FS_DIRTY_NODES);
1538 up_read(&sbi->node_write);
1539 unlock_page(page);
1540 return 0;
1541 }
1542
c9b60788 1543 if (__is_valid_data_blkaddr(ni.blk_addr) &&
89d13c38
JK
1544 !f2fs_is_valid_blkaddr(sbi, ni.blk_addr, DATA_GENERIC)) {
1545 up_read(&sbi->node_write);
c9b60788 1546 goto redirty_out;
89d13c38 1547 }
c9b60788 1548
e7c75ab0
JK
1549 if (atomic && !test_opt(sbi, NOBARRIER))
1550 fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
1551
faa24895 1552 set_page_writeback(page);
17c50035 1553 ClearPageError(page);
50fa53ec
CY
1554
1555 if (f2fs_in_warm_node_list(sbi, page)) {
1556 seq = f2fs_add_fsync_node_entry(sbi, page);
1557 if (seq_id)
1558 *seq_id = seq;
1559 }
1560
faa24895 1561 fio.old_blkaddr = ni.blk_addr;
4d57b86d 1562 f2fs_do_write_node_page(nid, &fio);
faa24895
JK
1563 set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
1564 dec_page_count(sbi, F2FS_DIRTY_NODES);
1565 up_read(&sbi->node_write);
1566
d68f735b 1567 if (wbc->for_reclaim) {
bab475c5 1568 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, NODE);
d68f735b
JK
1569 submitted = NULL;
1570 }
faa24895
JK
1571
1572 unlock_page(page);
1573
d68f735b 1574 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 1575 f2fs_submit_merged_write(sbi, NODE);
d68f735b
JK
1576 submitted = NULL;
1577 }
1578 if (submitted)
1579 *submitted = fio.submitted;
faa24895 1580
401db79f
YS
1581 if (do_balance)
1582 f2fs_balance_fs(sbi, false);
faa24895
JK
1583 return 0;
1584
1585redirty_out:
1586 redirty_page_for_writepage(wbc, page);
1587 return AOP_WRITEPAGE_ACTIVATE;
1588}
1589
48018b4c 1590int f2fs_move_node_page(struct page *node_page, int gc_type)
f15194fc 1591{
48018b4c
CY
1592 int err = 0;
1593
f15194fc
YH
1594 if (gc_type == FG_GC) {
1595 struct writeback_control wbc = {
1596 .sync_mode = WB_SYNC_ALL,
1597 .nr_to_write = 1,
1598 .for_reclaim = 0,
1599 };
1600
f15194fc 1601 f2fs_wait_on_page_writeback(node_page, NODE, true);
f15194fc 1602 f2fs_bug_on(F2FS_P_SB(node_page), PageWriteback(node_page));
8d64d365
CY
1603
1604 set_page_dirty(node_page);
1605
48018b4c
CY
1606 if (!clear_page_dirty_for_io(node_page)) {
1607 err = -EAGAIN;
f15194fc 1608 goto out_page;
48018b4c 1609 }
f15194fc
YH
1610
1611 if (__write_node_page(node_page, false, NULL,
48018b4c
CY
1612 &wbc, false, FS_GC_NODE_IO, NULL)) {
1613 err = -EAGAIN;
f15194fc 1614 unlock_page(node_page);
48018b4c 1615 }
f15194fc
YH
1616 goto release_page;
1617 } else {
1618 /* set page dirty and write it */
1619 if (!PageWriteback(node_page))
1620 set_page_dirty(node_page);
1621 }
1622out_page:
1623 unlock_page(node_page);
1624release_page:
1625 f2fs_put_page(node_page, 0);
48018b4c 1626 return err;
f15194fc
YH
1627}
1628
faa24895
JK
1629static int f2fs_write_node_page(struct page *page,
1630 struct writeback_control *wbc)
1631{
50fa53ec
CY
1632 return __write_node_page(page, false, NULL, wbc, false,
1633 FS_NODE_IO, NULL);
faa24895
JK
1634}
1635
4d57b86d 1636int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode,
50fa53ec
CY
1637 struct writeback_control *wbc, bool atomic,
1638 unsigned int *seq_id)
e05df3b1 1639{
028a63a6 1640 pgoff_t index;
e05df3b1 1641 struct pagevec pvec;
c267ec15 1642 int ret = 0;
608514de
JK
1643 struct page *last_page = NULL;
1644 bool marked = false;
26de9b11 1645 nid_t ino = inode->i_ino;
028a63a6 1646 int nr_pages;
bab475c5 1647 int nwritten = 0;
52681375 1648
608514de
JK
1649 if (atomic) {
1650 last_page = last_fsync_dnode(sbi, ino);
1651 if (IS_ERR_OR_NULL(last_page))
1652 return PTR_ERR_OR_ZERO(last_page);
1653 }
1654retry:
86679820 1655 pagevec_init(&pvec);
52681375 1656 index = 0;
028a63a6
JK
1657
1658 while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
67fd707f 1659 PAGECACHE_TAG_DIRTY))) {
028a63a6 1660 int i;
52681375
JK
1661
1662 for (i = 0; i < nr_pages; i++) {
1663 struct page *page = pvec.pages[i];
d68f735b 1664 bool submitted = false;
52681375
JK
1665
1666 if (unlikely(f2fs_cp_error(sbi))) {
608514de 1667 f2fs_put_page(last_page, 0);
52681375 1668 pagevec_release(&pvec);
9de69279
CY
1669 ret = -EIO;
1670 goto out;
52681375
JK
1671 }
1672
1673 if (!IS_DNODE(page) || !is_cold_node(page))
1674 continue;
1675 if (ino_of_node(page) != ino)
1676 continue;
1677
1678 lock_page(page);
1679
1680 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1681continue_unlock:
1682 unlock_page(page);
1683 continue;
1684 }
1685 if (ino_of_node(page) != ino)
1686 goto continue_unlock;
1687
608514de 1688 if (!PageDirty(page) && page != last_page) {
52681375
JK
1689 /* someone wrote it for us */
1690 goto continue_unlock;
1691 }
1692
1693 f2fs_wait_on_page_writeback(page, NODE, true);
1694 BUG_ON(PageWriteback(page));
52681375 1695
d29fd172
JK
1696 set_fsync_mark(page, 0);
1697 set_dentry_mark(page, 0);
1698
608514de
JK
1699 if (!atomic || page == last_page) {
1700 set_fsync_mark(page, 1);
26de9b11
JK
1701 if (IS_INODE(page)) {
1702 if (is_inode_flag_set(inode,
1703 FI_DIRTY_INODE))
4d57b86d 1704 f2fs_update_inode(inode, page);
608514de 1705 set_dentry_mark(page,
4d57b86d 1706 f2fs_need_dentry_mark(sbi, ino));
26de9b11 1707 }
608514de
JK
1708 /* may be written by other thread */
1709 if (!PageDirty(page))
1710 set_page_dirty(page);
1711 }
1712
1713 if (!clear_page_dirty_for_io(page))
1714 goto continue_unlock;
52681375 1715
e7c75ab0 1716 ret = __write_node_page(page, atomic &&
d68f735b 1717 page == last_page,
b0af6d49 1718 &submitted, wbc, true,
50fa53ec 1719 FS_NODE_IO, seq_id);
c267ec15 1720 if (ret) {
52681375 1721 unlock_page(page);
608514de
JK
1722 f2fs_put_page(last_page, 0);
1723 break;
d68f735b 1724 } else if (submitted) {
bab475c5 1725 nwritten++;
608514de 1726 }
3f5f4959 1727
608514de
JK
1728 if (page == last_page) {
1729 f2fs_put_page(page, 0);
1730 marked = true;
52681375 1731 break;
c267ec15 1732 }
52681375
JK
1733 }
1734 pagevec_release(&pvec);
1735 cond_resched();
1736
608514de 1737 if (ret || marked)
52681375
JK
1738 break;
1739 }
608514de
JK
1740 if (!ret && atomic && !marked) {
1741 f2fs_msg(sbi->sb, KERN_DEBUG,
1742 "Retry to write fsync mark: ino=%u, idx=%lx",
1743 ino, last_page->index);
1744 lock_page(last_page);
d40a43af 1745 f2fs_wait_on_page_writeback(last_page, NODE, true);
608514de
JK
1746 set_page_dirty(last_page);
1747 unlock_page(last_page);
1748 goto retry;
1749 }
9de69279 1750out:
bab475c5
CY
1751 if (nwritten)
1752 f2fs_submit_merged_write_cond(sbi, NULL, NULL, ino, NODE);
c267ec15 1753 return ret ? -EIO: 0;
52681375
JK
1754}
1755
4d57b86d
CY
1756int f2fs_sync_node_pages(struct f2fs_sb_info *sbi,
1757 struct writeback_control *wbc,
b0af6d49 1758 bool do_balance, enum iostat_type io_type)
52681375 1759{
028a63a6 1760 pgoff_t index;
52681375
JK
1761 struct pagevec pvec;
1762 int step = 0;
12bb0a8f 1763 int nwritten = 0;
3f5f4959 1764 int ret = 0;
c29fd0c0 1765 int nr_pages, done = 0;
e05df3b1 1766
86679820 1767 pagevec_init(&pvec);
e05df3b1
JK
1768
1769next_step:
1770 index = 0;
028a63a6 1771
c29fd0c0
CY
1772 while (!done && (nr_pages = pagevec_lookup_tag(&pvec,
1773 NODE_MAPPING(sbi), &index, PAGECACHE_TAG_DIRTY))) {
028a63a6 1774 int i;
e05df3b1
JK
1775
1776 for (i = 0; i < nr_pages; i++) {
1777 struct page *page = pvec.pages[i];
d68f735b 1778 bool submitted = false;
e05df3b1 1779
c29fd0c0
CY
1780 /* give a priority to WB_SYNC threads */
1781 if (atomic_read(&sbi->wb_sync_req[NODE]) &&
1782 wbc->sync_mode == WB_SYNC_NONE) {
1783 done = 1;
1784 break;
1785 }
1786
e05df3b1
JK
1787 /*
1788 * flushing sequence with step:
1789 * 0. indirect nodes
1790 * 1. dentry dnodes
1791 * 2. file dnodes
1792 */
1793 if (step == 0 && IS_DNODE(page))
1794 continue;
1795 if (step == 1 && (!IS_DNODE(page) ||
1796 is_cold_node(page)))
1797 continue;
1798 if (step == 2 && (!IS_DNODE(page) ||
1799 !is_cold_node(page)))
1800 continue;
9a4cbc9e 1801lock_node:
4b270a8c
CY
1802 if (wbc->sync_mode == WB_SYNC_ALL)
1803 lock_page(page);
1804 else if (!trylock_page(page))
e05df3b1
JK
1805 continue;
1806
4ef51a8f 1807 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
e05df3b1
JK
1808continue_unlock:
1809 unlock_page(page);
1810 continue;
1811 }
e05df3b1
JK
1812
1813 if (!PageDirty(page)) {
1814 /* someone wrote it for us */
1815 goto continue_unlock;
1816 }
1817
2049d4fc 1818 /* flush inline_data */
52681375 1819 if (is_inline_node(page)) {
2049d4fc
JK
1820 clear_inline_node(page);
1821 unlock_page(page);
1822 flush_inline_data(sbi, ino_of_node(page));
9a4cbc9e 1823 goto lock_node;
2049d4fc
JK
1824 }
1825
fa3d2bdf
JK
1826 f2fs_wait_on_page_writeback(page, NODE, true);
1827
1828 BUG_ON(PageWriteback(page));
e05df3b1
JK
1829 if (!clear_page_dirty_for_io(page))
1830 goto continue_unlock;
1831
52681375
JK
1832 set_fsync_mark(page, 0);
1833 set_dentry_mark(page, 0);
52746519 1834
401db79f 1835 ret = __write_node_page(page, false, &submitted,
50fa53ec 1836 wbc, do_balance, io_type, NULL);
d68f735b 1837 if (ret)
52746519 1838 unlock_page(page);
d68f735b 1839 else if (submitted)
3f5f4959 1840 nwritten++;
e05df3b1
JK
1841
1842 if (--wbc->nr_to_write == 0)
1843 break;
1844 }
1845 pagevec_release(&pvec);
1846 cond_resched();
1847
1848 if (wbc->nr_to_write == 0) {
1849 step = 2;
1850 break;
1851 }
1852 }
1853
1854 if (step < 2) {
fd8c8caf
CY
1855 if (wbc->sync_mode == WB_SYNC_NONE && step == 1)
1856 goto out;
e05df3b1
JK
1857 step++;
1858 goto next_step;
1859 }
fd8c8caf 1860out:
3f5f4959 1861 if (nwritten)
b9109b0e 1862 f2fs_submit_merged_write(sbi, NODE);
db198ae0
CY
1863
1864 if (unlikely(f2fs_cp_error(sbi)))
1865 return -EIO;
3f5f4959 1866 return ret;
e05df3b1
JK
1867}
1868
50fa53ec
CY
1869int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi,
1870 unsigned int seq_id)
cfe58f9d 1871{
50fa53ec
CY
1872 struct fsync_node_entry *fn;
1873 struct page *page;
1874 struct list_head *head = &sbi->fsync_node_list;
1875 unsigned long flags;
1876 unsigned int cur_seq_id = 0;
280db3c8 1877 int ret2, ret = 0;
cfe58f9d 1878
50fa53ec
CY
1879 while (seq_id && cur_seq_id < seq_id) {
1880 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
1881 if (list_empty(head)) {
1882 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
1883 break;
1884 }
1885 fn = list_first_entry(head, struct fsync_node_entry, list);
1886 if (fn->seq_id > seq_id) {
1887 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
1888 break;
1889 }
1890 cur_seq_id = fn->seq_id;
1891 page = fn->page;
1892 get_page(page);
1893 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
4ef51a8f 1894
50fa53ec
CY
1895 f2fs_wait_on_page_writeback(page, NODE, true);
1896 if (TestClearPageError(page))
1897 ret = -EIO;
cfe58f9d 1898
50fa53ec 1899 put_page(page);
cfe58f9d 1900
50fa53ec
CY
1901 if (ret)
1902 break;
cfe58f9d
JK
1903 }
1904
280db3c8 1905 ret2 = filemap_check_errors(NODE_MAPPING(sbi));
cfe58f9d
JK
1906 if (!ret)
1907 ret = ret2;
50fa53ec 1908
cfe58f9d
JK
1909 return ret;
1910}
1911
e05df3b1
JK
1912static int f2fs_write_node_pages(struct address_space *mapping,
1913 struct writeback_control *wbc)
1914{
4081363f 1915 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
9dfa1baf 1916 struct blk_plug plug;
50c8cdb3 1917 long diff;
e05df3b1 1918
0771fcc7
CY
1919 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1920 goto skip_write;
1921
4660f9c0
JK
1922 /* balancing f2fs's metadata in background */
1923 f2fs_balance_fs_bg(sbi);
e05df3b1 1924
a7fdffbd 1925 /* collect a number of dirty node pages and write together */
87d6f890 1926 if (get_pages(sbi, F2FS_DIRTY_NODES) < nr_pages_to_skip(sbi, NODE))
d3baf95d 1927 goto skip_write;
a7fdffbd 1928
c29fd0c0
CY
1929 if (wbc->sync_mode == WB_SYNC_ALL)
1930 atomic_inc(&sbi->wb_sync_req[NODE]);
1931 else if (atomic_read(&sbi->wb_sync_req[NODE]))
1932 goto skip_write;
1933
d31c7c3f
YH
1934 trace_f2fs_writepages(mapping->host, wbc, NODE);
1935
50c8cdb3 1936 diff = nr_pages_to_write(sbi, NODE, wbc);
9dfa1baf 1937 blk_start_plug(&plug);
4d57b86d 1938 f2fs_sync_node_pages(sbi, wbc, true, FS_NODE_IO);
9dfa1baf 1939 blk_finish_plug(&plug);
50c8cdb3 1940 wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
c29fd0c0
CY
1941
1942 if (wbc->sync_mode == WB_SYNC_ALL)
1943 atomic_dec(&sbi->wb_sync_req[NODE]);
e05df3b1 1944 return 0;
d3baf95d
JK
1945
1946skip_write:
1947 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES);
d31c7c3f 1948 trace_f2fs_writepages(mapping->host, wbc, NODE);
d3baf95d 1949 return 0;
e05df3b1
JK
1950}
1951
1952static int f2fs_set_node_page_dirty(struct page *page)
1953{
26c6b887
JK
1954 trace_f2fs_set_page_dirty(page, NODE);
1955
237c0790
JK
1956 if (!PageUptodate(page))
1957 SetPageUptodate(page);
54c55c4e
WG
1958#ifdef CONFIG_F2FS_CHECK_FS
1959 if (IS_INODE(page))
1960 f2fs_inode_chksum_set(F2FS_P_SB(page), page);
1961#endif
e05df3b1 1962 if (!PageDirty(page)) {
b87078ad 1963 __set_page_dirty_nobuffers(page);
4081363f 1964 inc_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
e05df3b1 1965 SetPagePrivate(page);
9e4ded3f 1966 f2fs_trace_pid(page);
e05df3b1
JK
1967 return 1;
1968 }
1969 return 0;
1970}
1971
0a8165d7 1972/*
e05df3b1
JK
1973 * Structure of the f2fs node operations
1974 */
1975const struct address_space_operations f2fs_node_aops = {
1976 .writepage = f2fs_write_node_page,
1977 .writepages = f2fs_write_node_pages,
1978 .set_page_dirty = f2fs_set_node_page_dirty,
487261f3
CY
1979 .invalidatepage = f2fs_invalidate_page,
1980 .releasepage = f2fs_release_page,
5b7a487c
WG
1981#ifdef CONFIG_MIGRATION
1982 .migratepage = f2fs_migrate_page,
1983#endif
e05df3b1
JK
1984};
1985
8a7ed66a
JK
1986static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i,
1987 nid_t n)
e05df3b1 1988{
8a7ed66a 1989 return radix_tree_lookup(&nm_i->free_nid_root, n);
e05df3b1
JK
1990}
1991
9a4ffdf5 1992static int __insert_free_nid(struct f2fs_sb_info *sbi,
a0761f63 1993 struct free_nid *i, enum nid_state state)
e05df3b1 1994{
b8559dc2
CY
1995 struct f2fs_nm_info *nm_i = NM_I(sbi);
1996
a0761f63
FL
1997 int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i);
1998 if (err)
1999 return err;
eb0aa4b8 2000
9a4ffdf5
CY
2001 f2fs_bug_on(sbi, state != i->state);
2002 nm_i->nid_cnt[state]++;
2003 if (state == FREE_NID)
2004 list_add_tail(&i->list, &nm_i->free_nid_list);
eb0aa4b8 2005 return 0;
b8559dc2
CY
2006}
2007
9a4ffdf5 2008static void __remove_free_nid(struct f2fs_sb_info *sbi,
a0761f63 2009 struct free_nid *i, enum nid_state state)
b8559dc2
CY
2010{
2011 struct f2fs_nm_info *nm_i = NM_I(sbi);
2012
9a4ffdf5
CY
2013 f2fs_bug_on(sbi, state != i->state);
2014 nm_i->nid_cnt[state]--;
2015 if (state == FREE_NID)
2016 list_del(&i->list);
a0761f63
FL
2017 radix_tree_delete(&nm_i->free_nid_root, i->nid);
2018}
2019
2020static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i,
2021 enum nid_state org_state, enum nid_state dst_state)
b8559dc2
CY
2022{
2023 struct f2fs_nm_info *nm_i = NM_I(sbi);
2024
a0761f63
FL
2025 f2fs_bug_on(sbi, org_state != i->state);
2026 i->state = dst_state;
2027 nm_i->nid_cnt[org_state]--;
2028 nm_i->nid_cnt[dst_state]++;
2029
2030 switch (dst_state) {
2031 case PREALLOC_NID:
2032 list_del(&i->list);
2033 break;
2034 case FREE_NID:
2035 list_add_tail(&i->list, &nm_i->free_nid_list);
2036 break;
2037 default:
2038 BUG_ON(1);
2039 }
e05df3b1
JK
2040}
2041
5921aaa1
L
2042static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid,
2043 bool set, bool build)
2044{
2045 struct f2fs_nm_info *nm_i = NM_I(sbi);
2046 unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid);
2047 unsigned int nid_ofs = nid - START_NID(nid);
2048
2049 if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap))
2050 return;
2051
2052 if (set) {
2053 if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2054 return;
2055 __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2056 nm_i->free_nid_count[nat_ofs]++;
2057 } else {
2058 if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2059 return;
2060 __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2061 if (!build)
2062 nm_i->free_nid_count[nat_ofs]--;
2063 }
2064}
2065
4ac91242 2066/* return if the nid is recognized as free */
5921aaa1
L
2067static bool add_free_nid(struct f2fs_sb_info *sbi,
2068 nid_t nid, bool build, bool update)
e05df3b1 2069{
6fb03f3a 2070 struct f2fs_nm_info *nm_i = NM_I(sbi);
30a61ddf 2071 struct free_nid *i, *e;
59bbd474 2072 struct nat_entry *ne;
30a61ddf
CY
2073 int err = -EINVAL;
2074 bool ret = false;
9198aceb
JK
2075
2076 /* 0 nid should not be used */
cfb271d4 2077 if (unlikely(nid == 0))
4ac91242 2078 return false;
59bbd474 2079
7bd59381 2080 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS);
e05df3b1 2081 i->nid = nid;
9a4ffdf5 2082 i->state = FREE_NID;
e05df3b1 2083
5921aaa1 2084 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
769ec6e5 2085
b8559dc2 2086 spin_lock(&nm_i->nid_list_lock);
30a61ddf
CY
2087
2088 if (build) {
2089 /*
2090 * Thread A Thread B
2091 * - f2fs_create
2092 * - f2fs_new_inode
4d57b86d 2093 * - f2fs_alloc_nid
9a4ffdf5 2094 * - __insert_nid_to_list(PREALLOC_NID)
30a61ddf 2095 * - f2fs_balance_fs_bg
4d57b86d
CY
2096 * - f2fs_build_free_nids
2097 * - __f2fs_build_free_nids
30a61ddf
CY
2098 * - scan_nat_page
2099 * - add_free_nid
2100 * - __lookup_nat_cache
2101 * - f2fs_add_link
4d57b86d
CY
2102 * - f2fs_init_inode_metadata
2103 * - f2fs_new_inode_page
2104 * - f2fs_new_node_page
30a61ddf 2105 * - set_node_addr
4d57b86d 2106 * - f2fs_alloc_nid_done
9a4ffdf5
CY
2107 * - __remove_nid_from_list(PREALLOC_NID)
2108 * - __insert_nid_to_list(FREE_NID)
30a61ddf
CY
2109 */
2110 ne = __lookup_nat_cache(nm_i, nid);
2111 if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) ||
2112 nat_get_blkaddr(ne) != NULL_ADDR))
2113 goto err_out;
2114
2115 e = __lookup_free_nid_list(nm_i, nid);
2116 if (e) {
9a4ffdf5 2117 if (e->state == FREE_NID)
30a61ddf
CY
2118 ret = true;
2119 goto err_out;
2120 }
2121 }
2122 ret = true;
a0761f63 2123 err = __insert_free_nid(sbi, i, FREE_NID);
30a61ddf 2124err_out:
5921aaa1
L
2125 if (update) {
2126 update_free_nid_bitmap(sbi, nid, ret, build);
2127 if (!build)
2128 nm_i->available_nids++;
2129 }
eb0aa4b8
JK
2130 spin_unlock(&nm_i->nid_list_lock);
2131 radix_tree_preload_end();
5921aaa1 2132
30a61ddf 2133 if (err)
e05df3b1 2134 kmem_cache_free(free_nid_slab, i);
30a61ddf 2135 return ret;
e05df3b1
JK
2136}
2137
b8559dc2 2138static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2139{
b8559dc2 2140 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1 2141 struct free_nid *i;
cf0ee0f0
CY
2142 bool need_free = false;
2143
b8559dc2 2144 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2145 i = __lookup_free_nid_list(nm_i, nid);
9a4ffdf5 2146 if (i && i->state == FREE_NID) {
a0761f63 2147 __remove_free_nid(sbi, i, FREE_NID);
cf0ee0f0 2148 need_free = true;
e05df3b1 2149 }
b8559dc2 2150 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY
2151
2152 if (need_free)
2153 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK
2154}
2155
e2374015 2156static int scan_nat_page(struct f2fs_sb_info *sbi,
e05df3b1
JK
2157 struct page *nat_page, nid_t start_nid)
2158{
6fb03f3a 2159 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1
JK
2160 struct f2fs_nat_block *nat_blk = page_address(nat_page);
2161 block_t blk_addr;
4ac91242 2162 unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid);
e05df3b1
JK
2163 int i;
2164
23380b85 2165 __set_bit_le(nat_ofs, nm_i->nat_block_bitmap);
4ac91242 2166
e05df3b1
JK
2167 i = start_nid % NAT_ENTRY_PER_BLOCK;
2168
2169 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
cfb271d4 2170 if (unlikely(start_nid >= nm_i->max_nid))
04431c44 2171 break;
23d38844
HL
2172
2173 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
e2374015
CY
2174
2175 if (blk_addr == NEW_ADDR)
2176 return -EINVAL;
2177
5921aaa1
L
2178 if (blk_addr == NULL_ADDR) {
2179 add_free_nid(sbi, start_nid, true, true);
2180 } else {
2181 spin_lock(&NM_I(sbi)->nid_list_lock);
2182 update_free_nid_bitmap(sbi, start_nid, false, true);
2183 spin_unlock(&NM_I(sbi)->nid_list_lock);
2184 }
4ac91242 2185 }
e2374015
CY
2186
2187 return 0;
4ac91242
CY
2188}
2189
2fbaa25f 2190static void scan_curseg_cache(struct f2fs_sb_info *sbi)
4ac91242 2191{
4ac91242
CY
2192 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2193 struct f2fs_journal *journal = curseg->journal;
2fbaa25f
CY
2194 int i;
2195
2196 down_read(&curseg->journal_rwsem);
2197 for (i = 0; i < nats_in_cursum(journal); i++) {
2198 block_t addr;
2199 nid_t nid;
2200
2201 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
2202 nid = le32_to_cpu(nid_in_journal(journal, i));
2203 if (addr == NULL_ADDR)
5921aaa1 2204 add_free_nid(sbi, nid, true, false);
2fbaa25f
CY
2205 else
2206 remove_free_nid(sbi, nid);
2207 }
2208 up_read(&curseg->journal_rwsem);
2209}
2210
2211static void scan_free_nid_bits(struct f2fs_sb_info *sbi)
2212{
2213 struct f2fs_nm_info *nm_i = NM_I(sbi);
4ac91242 2214 unsigned int i, idx;
97456574 2215 nid_t nid;
4ac91242
CY
2216
2217 down_read(&nm_i->nat_tree_lock);
2218
2219 for (i = 0; i < nm_i->nat_blocks; i++) {
2220 if (!test_bit_le(i, nm_i->nat_block_bitmap))
2221 continue;
586d1492
CY
2222 if (!nm_i->free_nid_count[i])
2223 continue;
4ac91242 2224 for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) {
97456574
FL
2225 idx = find_next_bit_le(nm_i->free_nid_bitmap[i],
2226 NAT_ENTRY_PER_BLOCK, idx);
2227 if (idx >= NAT_ENTRY_PER_BLOCK)
2228 break;
4ac91242
CY
2229
2230 nid = i * NAT_ENTRY_PER_BLOCK + idx;
5921aaa1 2231 add_free_nid(sbi, nid, true, false);
4ac91242 2232
9a4ffdf5 2233 if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS)
4ac91242
CY
2234 goto out;
2235 }
2236 }
2237out:
2fbaa25f 2238 scan_curseg_cache(sbi);
4ac91242 2239
4ac91242 2240 up_read(&nm_i->nat_tree_lock);
e05df3b1
JK
2241}
2242
e2374015 2243static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
4d57b86d 2244 bool sync, bool mount)
e05df3b1 2245{
e05df3b1 2246 struct f2fs_nm_info *nm_i = NM_I(sbi);
e2374015 2247 int i = 0, ret;
55008d84 2248 nid_t nid = nm_i->next_scan_nid;
e05df3b1 2249
e9cdd307
YH
2250 if (unlikely(nid >= nm_i->max_nid))
2251 nid = 0;
2252
55008d84 2253 /* Enough entries */
9a4ffdf5 2254 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2255 return 0;
e05df3b1 2256
4d57b86d 2257 if (!sync && !f2fs_available_free_memory(sbi, FREE_NIDS))
e2374015 2258 return 0;
e05df3b1 2259
4ac91242
CY
2260 if (!mount) {
2261 /* try to find free nids in free_nid_bitmap */
2262 scan_free_nid_bits(sbi);
2263
74986213 2264 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2265 return 0;
22ad0b6a
JK
2266 }
2267
55008d84 2268 /* readahead nat pages to be scanned */
4d57b86d 2269 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
26879fb1 2270 META_NAT, true);
e05df3b1 2271
b873b798 2272 down_read(&nm_i->nat_tree_lock);
a5131193 2273
e05df3b1 2274 while (1) {
66e83361
YH
2275 if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
2276 nm_i->nat_block_bitmap)) {
2277 struct page *page = get_current_nat_page(sbi, nid);
e05df3b1 2278
edc55aaf
JK
2279 if (IS_ERR(page)) {
2280 ret = PTR_ERR(page);
2281 } else {
2282 ret = scan_nat_page(sbi, page, nid);
2283 f2fs_put_page(page, 1);
2284 }
e2374015
CY
2285
2286 if (ret) {
2287 up_read(&nm_i->nat_tree_lock);
2288 f2fs_bug_on(sbi, !mount);
2289 f2fs_msg(sbi->sb, KERN_ERR,
2290 "NAT is corrupt, run fsck to fix it");
edc55aaf 2291 return ret;
e2374015 2292 }
66e83361 2293 }
e05df3b1
JK
2294
2295 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
cfb271d4 2296 if (unlikely(nid >= nm_i->max_nid))
e05df3b1 2297 nid = 0;
55008d84 2298
a6d494b6 2299 if (++i >= FREE_NID_PAGES)
e05df3b1
JK
2300 break;
2301 }
2302
55008d84
JK
2303 /* go to the next free nat pages to find free nids abundantly */
2304 nm_i->next_scan_nid = nid;
e05df3b1
JK
2305
2306 /* find free nids from current sum_pages */
2fbaa25f 2307 scan_curseg_cache(sbi);
dfc08a12 2308
b873b798 2309 up_read(&nm_i->nat_tree_lock);
2db2388f 2310
4d57b86d 2311 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
ea1a29a0 2312 nm_i->ra_nid_pages, META_NAT, false);
e2374015
CY
2313
2314 return 0;
e05df3b1
JK
2315}
2316
e2374015 2317int f2fs_build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
2411cf5b 2318{
e2374015
CY
2319 int ret;
2320
2411cf5b 2321 mutex_lock(&NM_I(sbi)->build_lock);
e2374015 2322 ret = __f2fs_build_free_nids(sbi, sync, mount);
2411cf5b 2323 mutex_unlock(&NM_I(sbi)->build_lock);
e2374015
CY
2324
2325 return ret;
2411cf5b
CY
2326}
2327
e05df3b1
JK
2328/*
2329 * If this function returns success, caller can obtain a new nid
2330 * from second parameter of this function.
2331 * The returned nid could be used ino as well as nid when inode is created.
2332 */
4d57b86d 2333bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
e05df3b1
JK
2334{
2335 struct f2fs_nm_info *nm_i = NM_I(sbi);
2336 struct free_nid *i = NULL;
e05df3b1 2337retry:
55523519
CY
2338 if (time_to_inject(sbi, FAULT_ALLOC_NID)) {
2339 f2fs_show_injection_info(FAULT_ALLOC_NID);
cb78942b 2340 return false;
55523519 2341 }
7fa750a1 2342
b8559dc2 2343 spin_lock(&nm_i->nid_list_lock);
e05df3b1 2344
04d47e67
CY
2345 if (unlikely(nm_i->available_nids == 0)) {
2346 spin_unlock(&nm_i->nid_list_lock);
2347 return false;
2348 }
e05df3b1 2349
4d57b86d
CY
2350 /* We should not use stale free nids created by f2fs_build_free_nids */
2351 if (nm_i->nid_cnt[FREE_NID] && !on_f2fs_build_free_nids(nm_i)) {
9a4ffdf5
CY
2352 f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list));
2353 i = list_first_entry(&nm_i->free_nid_list,
b8559dc2 2354 struct free_nid, list);
55008d84 2355 *nid = i->nid;
b8559dc2 2356
a0761f63 2357 __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID);
04d47e67 2358 nm_i->available_nids--;
4ac91242 2359
346fe752 2360 update_free_nid_bitmap(sbi, *nid, false, false);
4ac91242 2361
b8559dc2 2362 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK
2363 return true;
2364 }
b8559dc2 2365 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK
2366
2367 /* Let's scan nat pages and its caches to get free nids */
f84262b0
JK
2368 if (!f2fs_build_free_nids(sbi, true, false))
2369 goto retry;
2370 return false;
e05df3b1
JK
2371}
2372
0a8165d7 2373/*
4d57b86d 2374 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2375 */
4d57b86d 2376void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK
2377{
2378 struct f2fs_nm_info *nm_i = NM_I(sbi);
2379 struct free_nid *i;
2380
b8559dc2 2381 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2382 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2 2383 f2fs_bug_on(sbi, !i);
a0761f63 2384 __remove_free_nid(sbi, i, PREALLOC_NID);
b8559dc2 2385 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY
2386
2387 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK
2388}
2389
0a8165d7 2390/*
4d57b86d 2391 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2392 */
4d57b86d 2393void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2394{
49952fa1
JK
2395 struct f2fs_nm_info *nm_i = NM_I(sbi);
2396 struct free_nid *i;
cf0ee0f0 2397 bool need_free = false;
49952fa1 2398
65985d93
JK
2399 if (!nid)
2400 return;
2401
b8559dc2 2402 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2403 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2
CY
2404 f2fs_bug_on(sbi, !i);
2405
4d57b86d 2406 if (!f2fs_available_free_memory(sbi, FREE_NIDS)) {
a0761f63 2407 __remove_free_nid(sbi, i, PREALLOC_NID);
cf0ee0f0 2408 need_free = true;
95630cba 2409 } else {
a0761f63 2410 __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID);
95630cba 2411 }
04d47e67
CY
2412
2413 nm_i->available_nids++;
2414
346fe752 2415 update_free_nid_bitmap(sbi, nid, true, false);
4ac91242 2416
b8559dc2 2417 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY
2418
2419 if (need_free)
2420 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK
2421}
2422
4d57b86d 2423int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
31696580
CY
2424{
2425 struct f2fs_nm_info *nm_i = NM_I(sbi);
2426 struct free_nid *i, *next;
2427 int nr = nr_shrink;
2428
9a4ffdf5 2429 if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
ad4edb83
JK
2430 return 0;
2431
31696580
CY
2432 if (!mutex_trylock(&nm_i->build_lock))
2433 return 0;
2434
b8559dc2 2435 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 2436 list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) {
b8559dc2 2437 if (nr_shrink <= 0 ||
9a4ffdf5 2438 nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
31696580 2439 break;
b8559dc2 2440
a0761f63 2441 __remove_free_nid(sbi, i, FREE_NID);
31696580
CY
2442 kmem_cache_free(free_nid_slab, i);
2443 nr_shrink--;
31696580 2444 }
b8559dc2 2445 spin_unlock(&nm_i->nid_list_lock);
31696580
CY
2446 mutex_unlock(&nm_i->build_lock);
2447
2448 return nr - nr_shrink;
2449}
2450
4d57b86d 2451void f2fs_recover_inline_xattr(struct inode *inode, struct page *page)
28cdce04 2452{
28cdce04
CY
2453 void *src_addr, *dst_addr;
2454 size_t inline_size;
2455 struct page *ipage;
2456 struct f2fs_inode *ri;
2457
4d57b86d 2458 ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino);
9850cf4a 2459 f2fs_bug_on(F2FS_I_SB(inode), IS_ERR(ipage));
28cdce04 2460
e3b4d43f 2461 ri = F2FS_INODE(page);
1eca05aa
YH
2462 if (ri->i_inline & F2FS_INLINE_XATTR) {
2463 set_inode_flag(inode, FI_INLINE_XATTR);
2464 } else {
91942321 2465 clear_inode_flag(inode, FI_INLINE_XATTR);
e3b4d43f
JK
2466 goto update_inode;
2467 }
2468
6afc662e
CY
2469 dst_addr = inline_xattr_addr(inode, ipage);
2470 src_addr = inline_xattr_addr(inode, page);
28cdce04
CY
2471 inline_size = inline_xattr_size(inode);
2472
fec1d657 2473 f2fs_wait_on_page_writeback(ipage, NODE, true);
28cdce04 2474 memcpy(dst_addr, src_addr, inline_size);
e3b4d43f 2475update_inode:
4d57b86d 2476 f2fs_update_inode(inode, ipage);
28cdce04
CY
2477 f2fs_put_page(ipage, 1);
2478}
2479
4d57b86d 2480int f2fs_recover_xattr_data(struct inode *inode, struct page *page)
abb2366c 2481{
4081363f 2482 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
abb2366c 2483 nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
87905682
YH
2484 nid_t new_xnid;
2485 struct dnode_of_data dn;
abb2366c 2486 struct node_info ni;
d260081c 2487 struct page *xpage;
7735730d 2488 int err;
abb2366c 2489
abb2366c
JK
2490 if (!prev_xnid)
2491 goto recover_xnid;
2492
d260081c 2493 /* 1: invalidate the previous xattr nid */
7735730d
CY
2494 err = f2fs_get_node_info(sbi, prev_xnid, &ni);
2495 if (err)
2496 return err;
2497
4d57b86d 2498 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 2499 dec_valid_node_count(sbi, inode, false);
479f40c4 2500 set_node_addr(sbi, &ni, NULL_ADDR, false);
abb2366c
JK
2501
2502recover_xnid:
d260081c 2503 /* 2: update xattr nid in inode */
4d57b86d 2504 if (!f2fs_alloc_nid(sbi, &new_xnid))
87905682
YH
2505 return -ENOSPC;
2506
2507 set_new_dnode(&dn, inode, NULL, NULL, new_xnid);
4d57b86d 2508 xpage = f2fs_new_node_page(&dn, XATTR_NODE_OFFSET);
87905682 2509 if (IS_ERR(xpage)) {
4d57b86d 2510 f2fs_alloc_nid_failed(sbi, new_xnid);
87905682
YH
2511 return PTR_ERR(xpage);
2512 }
2513
4d57b86d
CY
2514 f2fs_alloc_nid_done(sbi, new_xnid);
2515 f2fs_update_inode_page(inode);
d260081c
CY
2516
2517 /* 3: update and set xattr node page dirty */
87905682 2518 memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE);
abb2366c 2519
d260081c
CY
2520 set_page_dirty(xpage);
2521 f2fs_put_page(xpage, 1);
abb2366c 2522
d260081c 2523 return 0;
abb2366c
JK
2524}
2525
4d57b86d 2526int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
e05df3b1 2527{
58bfaf44 2528 struct f2fs_inode *src, *dst;
e05df3b1
JK
2529 nid_t ino = ino_of_node(page);
2530 struct node_info old_ni, new_ni;
2531 struct page *ipage;
7735730d 2532 int err;
e05df3b1 2533
7735730d
CY
2534 err = f2fs_get_node_info(sbi, ino, &old_ni);
2535 if (err)
2536 return err;
e8271fa3
JK
2537
2538 if (unlikely(old_ni.blk_addr != NULL_ADDR))
2539 return -EINVAL;
e8ea9b3d 2540retry:
300e129c 2541 ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false);
e8ea9b3d
JK
2542 if (!ipage) {
2543 congestion_wait(BLK_RW_ASYNC, HZ/50);
2544 goto retry;
2545 }
e05df3b1 2546
e1c42045 2547 /* Should not use this inode from free nid list */
b8559dc2 2548 remove_free_nid(sbi, ino);
e05df3b1 2549
237c0790
JK
2550 if (!PageUptodate(ipage))
2551 SetPageUptodate(ipage);
e05df3b1 2552 fill_node_footer(ipage, ino, ino, 0, true);
ef2a0071 2553 set_cold_node(ipage, false);
e05df3b1 2554
58bfaf44
JK
2555 src = F2FS_INODE(page);
2556 dst = F2FS_INODE(ipage);
e05df3b1 2557
58bfaf44
JK
2558 memcpy(dst, src, (unsigned long)&src->i_ext - (unsigned long)src);
2559 dst->i_size = 0;
2560 dst->i_blocks = cpu_to_le64(1);
2561 dst->i_links = cpu_to_le32(1);
2562 dst->i_xattr_nid = 0;
7a2af766 2563 dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR);
5c57132e 2564 if (dst->i_inline & F2FS_EXTRA_ATTR) {
7a2af766 2565 dst->i_extra_isize = src->i_extra_isize;
6afc662e 2566
7beb01f7 2567 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
6afc662e
CY
2568 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2569 i_inline_xattr_size))
2570 dst->i_inline_xattr_size = src->i_inline_xattr_size;
2571
7beb01f7 2572 if (f2fs_sb_has_project_quota(sbi) &&
5c57132e
CY
2573 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2574 i_projid))
2575 dst->i_projid = src->i_projid;
5cd1f387 2576
7beb01f7 2577 if (f2fs_sb_has_inode_crtime(sbi) &&
5cd1f387
CY
2578 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2579 i_crtime_nsec)) {
2580 dst->i_crtime = src->i_crtime;
2581 dst->i_crtime_nsec = src->i_crtime_nsec;
2582 }
5c57132e 2583 }
e05df3b1
JK
2584
2585 new_ni = old_ni;
2586 new_ni.ino = ino;
2587
0abd675e 2588 if (unlikely(inc_valid_node_count(sbi, NULL, true)))
65e5cd0a 2589 WARN_ON(1);
479f40c4 2590 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
e05df3b1 2591 inc_valid_inode_count(sbi);
617deb8c 2592 set_page_dirty(ipage);
e05df3b1
JK
2593 f2fs_put_page(ipage, 1);
2594 return 0;
2595}
2596
7735730d 2597int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
e05df3b1
JK
2598 unsigned int segno, struct f2fs_summary_block *sum)
2599{
2600 struct f2fs_node *rn;
2601 struct f2fs_summary *sum_entry;
e05df3b1 2602 block_t addr;
9ecf4b80 2603 int i, idx, last_offset, nrpages;
e05df3b1
JK
2604
2605 /* scan the node segment */
2606 last_offset = sbi->blocks_per_seg;
2607 addr = START_BLOCK(sbi, segno);
2608 sum_entry = &sum->entries[0];
2609
9ecf4b80 2610 for (i = 0; i < last_offset; i += nrpages, addr += nrpages) {
664ba972 2611 nrpages = min(last_offset - i, BIO_MAX_PAGES);
393ff91f 2612
e1c42045 2613 /* readahead node pages */
4d57b86d 2614 f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true);
e05df3b1 2615
9ecf4b80 2616 for (idx = addr; idx < addr + nrpages; idx++) {
4d57b86d 2617 struct page *page = f2fs_get_tmp_page(sbi, idx);
9af0ff1c 2618
7735730d
CY
2619 if (IS_ERR(page))
2620 return PTR_ERR(page);
2621
9ecf4b80
CY
2622 rn = F2FS_NODE(page);
2623 sum_entry->nid = rn->footer.nid;
2624 sum_entry->version = 0;
2625 sum_entry->ofs_in_node = 0;
2626 sum_entry++;
2627 f2fs_put_page(page, 1);
9af0ff1c 2628 }
bac4eef6 2629
9ecf4b80 2630 invalidate_mapping_pages(META_MAPPING(sbi), addr,
bac4eef6 2631 addr + nrpages);
e05df3b1 2632 }
7735730d 2633 return 0;
e05df3b1
JK
2634}
2635
aec71382 2636static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
e05df3b1
JK
2637{
2638 struct f2fs_nm_info *nm_i = NM_I(sbi);
2639 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2640 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK
2641 int i;
2642
b7ad7512 2643 down_write(&curseg->journal_rwsem);
dfc08a12 2644 for (i = 0; i < nats_in_cursum(journal); i++) {
e05df3b1
JK
2645 struct nat_entry *ne;
2646 struct f2fs_nat_entry raw_ne;
dfc08a12 2647 nid_t nid = le32_to_cpu(nid_in_journal(journal, i));
e05df3b1 2648
dfc08a12 2649 raw_ne = nat_in_journal(journal, i);
9be32d72 2650
e05df3b1 2651 ne = __lookup_nat_cache(nm_i, nid);
e05df3b1 2652 if (!ne) {
12f9ef37
YH
2653 ne = __alloc_nat_entry(nid, true);
2654 __init_nat_entry(nm_i, ne, &raw_ne, true);
e05df3b1 2655 }
04d47e67
CY
2656
2657 /*
2658 * if a free nat in journal has not been used after last
2659 * checkpoint, we should remove it from available nids,
2660 * since later we will add it again.
2661 */
2662 if (!get_nat_flag(ne, IS_DIRTY) &&
2663 le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) {
2664 spin_lock(&nm_i->nid_list_lock);
2665 nm_i->available_nids--;
2666 spin_unlock(&nm_i->nid_list_lock);
2667 }
2668
e05df3b1 2669 __set_nat_cache_dirty(nm_i, ne);
e05df3b1 2670 }
dfc08a12 2671 update_nats_in_cursum(journal, -i);
b7ad7512 2672 up_write(&curseg->journal_rwsem);
e05df3b1
JK
2673}
2674
309cc2b6
JK
2675static void __adjust_nat_entry_set(struct nat_entry_set *nes,
2676 struct list_head *head, int max)
e05df3b1 2677{
309cc2b6 2678 struct nat_entry_set *cur;
e05df3b1 2679
309cc2b6
JK
2680 if (nes->entry_cnt >= max)
2681 goto add_out;
e05df3b1 2682
309cc2b6
JK
2683 list_for_each_entry(cur, head, set_list) {
2684 if (cur->entry_cnt >= nes->entry_cnt) {
2685 list_add(&nes->set_list, cur->set_list.prev);
2686 return;
2687 }
aec71382 2688 }
309cc2b6
JK
2689add_out:
2690 list_add_tail(&nes->set_list, head);
2691}
e05df3b1 2692
9f7e4a2c 2693static void __update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
22ad0b6a
JK
2694 struct page *page)
2695{
2696 struct f2fs_nm_info *nm_i = NM_I(sbi);
2697 unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK;
2698 struct f2fs_nat_block *nat_blk = page_address(page);
2699 int valid = 0;
37a0ab2a 2700 int i = 0;
22ad0b6a
JK
2701
2702 if (!enabled_nat_bits(sbi, NULL))
2703 return;
2704
37a0ab2a
FL
2705 if (nat_index == 0) {
2706 valid = 1;
2707 i = 1;
2708 }
2709 for (; i < NAT_ENTRY_PER_BLOCK; i++) {
2710 if (nat_blk->entries[i].block_addr != NULL_ADDR)
22ad0b6a
JK
2711 valid++;
2712 }
2713 if (valid == 0) {
23380b85
JK
2714 __set_bit_le(nat_index, nm_i->empty_nat_bits);
2715 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK
2716 return;
2717 }
2718
23380b85 2719 __clear_bit_le(nat_index, nm_i->empty_nat_bits);
22ad0b6a 2720 if (valid == NAT_ENTRY_PER_BLOCK)
23380b85 2721 __set_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a 2722 else
23380b85 2723 __clear_bit_le(nat_index, nm_i->full_nat_bits);
22ad0b6a
JK
2724}
2725
edc55aaf 2726static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
22ad0b6a 2727 struct nat_entry_set *set, struct cp_control *cpc)
309cc2b6
JK
2728{
2729 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2730 struct f2fs_journal *journal = curseg->journal;
309cc2b6
JK
2731 nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK;
2732 bool to_journal = true;
2733 struct f2fs_nat_block *nat_blk;
2734 struct nat_entry *ne, *cur;
2735 struct page *page = NULL;
e05df3b1 2736
aec71382
CY
2737 /*
2738 * there are two steps to flush nat entries:
2739 * #1, flush nat entries to journal in current hot data summary block.
2740 * #2, flush nat entries to nat page.
2741 */
22ad0b6a
JK
2742 if (enabled_nat_bits(sbi, cpc) ||
2743 !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL))
309cc2b6
JK
2744 to_journal = false;
2745
2746 if (to_journal) {
b7ad7512 2747 down_write(&curseg->journal_rwsem);
309cc2b6
JK
2748 } else {
2749 page = get_next_nat_page(sbi, start_nid);
edc55aaf
JK
2750 if (IS_ERR(page))
2751 return PTR_ERR(page);
2752
309cc2b6
JK
2753 nat_blk = page_address(page);
2754 f2fs_bug_on(sbi, !nat_blk);
2755 }
aec71382 2756
309cc2b6
JK
2757 /* flush dirty nats in nat entry set */
2758 list_for_each_entry_safe(ne, cur, &set->entry_list, list) {
2759 struct f2fs_nat_entry *raw_ne;
2760 nid_t nid = nat_get_nid(ne);
2761 int offset;
2762
febeca6d 2763 f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR);
aec71382
CY
2764
2765 if (to_journal) {
4d57b86d 2766 offset = f2fs_lookup_journal_in_cursum(journal,
309cc2b6
JK
2767 NAT_JOURNAL, nid, 1);
2768 f2fs_bug_on(sbi, offset < 0);
dfc08a12
CY
2769 raw_ne = &nat_in_journal(journal, offset);
2770 nid_in_journal(journal, offset) = cpu_to_le32(nid);
aec71382 2771 } else {
309cc2b6 2772 raw_ne = &nat_blk->entries[nid - start_nid];
e05df3b1 2773 }
309cc2b6 2774 raw_nat_from_node_info(raw_ne, &ne->ni);
309cc2b6 2775 nat_reset_flag(ne);
0b28b71e 2776 __clear_nat_cache_dirty(NM_I(sbi), set, ne);
04d47e67 2777 if (nat_get_blkaddr(ne) == NULL_ADDR) {
5921aaa1 2778 add_free_nid(sbi, nid, false, true);
4ac91242
CY
2779 } else {
2780 spin_lock(&NM_I(sbi)->nid_list_lock);
346fe752 2781 update_free_nid_bitmap(sbi, nid, false, false);
04d47e67
CY
2782 spin_unlock(&NM_I(sbi)->nid_list_lock);
2783 }
309cc2b6 2784 }
e05df3b1 2785
22ad0b6a 2786 if (to_journal) {
b7ad7512 2787 up_write(&curseg->journal_rwsem);
22ad0b6a
JK
2788 } else {
2789 __update_nat_bits(sbi, start_nid, page);
309cc2b6 2790 f2fs_put_page(page, 1);
22ad0b6a 2791 }
aec71382 2792
59c9081b
YH
2793 /* Allow dirty nats by node block allocation in write_begin */
2794 if (!set->entry_cnt) {
2795 radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
2796 kmem_cache_free(nat_entry_set_slab, set);
2797 }
edc55aaf 2798 return 0;
309cc2b6 2799}
aec71382 2800
309cc2b6
JK
2801/*
2802 * This function is called during the checkpointing process.
2803 */
edc55aaf 2804int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
309cc2b6
JK
2805{
2806 struct f2fs_nm_info *nm_i = NM_I(sbi);
2807 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2808 struct f2fs_journal *journal = curseg->journal;
7aed0d45 2809 struct nat_entry_set *setvec[SETVEC_SIZE];
309cc2b6
JK
2810 struct nat_entry_set *set, *tmp;
2811 unsigned int found;
2812 nid_t set_idx = 0;
2813 LIST_HEAD(sets);
edc55aaf 2814 int err = 0;
309cc2b6 2815
7f2ecdd8
JK
2816 /* during unmount, let's flush nat_bits before checking dirty_nat_cnt */
2817 if (enabled_nat_bits(sbi, cpc)) {
2818 down_write(&nm_i->nat_tree_lock);
2819 remove_nats_in_journal(sbi);
2820 up_write(&nm_i->nat_tree_lock);
2821 }
2822
20d047c8 2823 if (!nm_i->dirty_nat_cnt)
edc55aaf 2824 return 0;
a5131193 2825
b873b798 2826 down_write(&nm_i->nat_tree_lock);
a5131193 2827
309cc2b6
JK
2828 /*
2829 * if there are no enough space in journal to store dirty nat
2830 * entries, remove all entries from journal and merge them
2831 * into nat entry set.
2832 */
900f7362 2833 if (enabled_nat_bits(sbi, cpc) ||
22ad0b6a 2834 !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL))
309cc2b6
JK
2835 remove_nats_in_journal(sbi);
2836
309cc2b6 2837 while ((found = __gang_lookup_nat_set(nm_i,
7aed0d45 2838 set_idx, SETVEC_SIZE, setvec))) {
309cc2b6
JK
2839 unsigned idx;
2840 set_idx = setvec[found - 1]->set + 1;
2841 for (idx = 0; idx < found; idx++)
2842 __adjust_nat_entry_set(setvec[idx], &sets,
dfc08a12 2843 MAX_NAT_JENTRIES(journal));
e05df3b1 2844 }
aec71382 2845
309cc2b6 2846 /* flush dirty nats in nat entry set */
edc55aaf
JK
2847 list_for_each_entry_safe(set, tmp, &sets, set_list) {
2848 err = __flush_nat_entry_set(sbi, set, cpc);
2849 if (err)
2850 break;
2851 }
309cc2b6 2852
b873b798 2853 up_write(&nm_i->nat_tree_lock);
59c9081b 2854 /* Allow dirty nats by node block allocation in write_begin */
edc55aaf
JK
2855
2856 return err;
e05df3b1
JK
2857}
2858
22ad0b6a
JK
2859static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
2860{
2861 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
2862 struct f2fs_nm_info *nm_i = NM_I(sbi);
2863 unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE;
2864 unsigned int i;
2865 __u64 cp_ver = cur_cp_version(ckpt);
22ad0b6a
JK
2866 block_t nat_bits_addr;
2867
2868 if (!enabled_nat_bits(sbi, NULL))
2869 return 0;
2870
df033caf 2871 nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
acbf054d
CY
2872 nm_i->nat_bits = f2fs_kzalloc(sbi,
2873 nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL);
22ad0b6a
JK
2874 if (!nm_i->nat_bits)
2875 return -ENOMEM;
2876
2877 nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg -
2878 nm_i->nat_bits_blocks;
2879 for (i = 0; i < nm_i->nat_bits_blocks; i++) {
7735730d
CY
2880 struct page *page;
2881
2882 page = f2fs_get_meta_page(sbi, nat_bits_addr++);
3b30eb19 2883 if (IS_ERR(page))
7735730d 2884 return PTR_ERR(page);
22ad0b6a
JK
2885
2886 memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS),
2887 page_address(page), F2FS_BLKSIZE);
2888 f2fs_put_page(page, 1);
2889 }
2890
ced2c7ea 2891 cp_ver |= (cur_cp_crc(ckpt) << 32);
22ad0b6a
JK
2892 if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) {
2893 disable_nat_bits(sbi, true);
2894 return 0;
2895 }
2896
2897 nm_i->full_nat_bits = nm_i->nat_bits + 8;
2898 nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes;
2899
2900 f2fs_msg(sbi->sb, KERN_NOTICE, "Found nat_bits in checkpoint");
2901 return 0;
2902}
2903
bd80a4b9 2904static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi)
7041d5d2
CY
2905{
2906 struct f2fs_nm_info *nm_i = NM_I(sbi);
2907 unsigned int i = 0;
2908 nid_t nid, last_nid;
2909
2910 if (!enabled_nat_bits(sbi, NULL))
2911 return;
2912
2913 for (i = 0; i < nm_i->nat_blocks; i++) {
2914 i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i);
2915 if (i >= nm_i->nat_blocks)
2916 break;
2917
2918 __set_bit_le(i, nm_i->nat_block_bitmap);
2919
2920 nid = i * NAT_ENTRY_PER_BLOCK;
f6986ede 2921 last_nid = nid + NAT_ENTRY_PER_BLOCK;
7041d5d2 2922
346fe752 2923 spin_lock(&NM_I(sbi)->nid_list_lock);
7041d5d2 2924 for (; nid < last_nid; nid++)
346fe752
CY
2925 update_free_nid_bitmap(sbi, nid, true, true);
2926 spin_unlock(&NM_I(sbi)->nid_list_lock);
7041d5d2
CY
2927 }
2928
2929 for (i = 0; i < nm_i->nat_blocks; i++) {
2930 i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i);
2931 if (i >= nm_i->nat_blocks)
2932 break;
2933
2934 __set_bit_le(i, nm_i->nat_block_bitmap);
2935 }
2936}
2937
e05df3b1
JK
2938static int init_node_manager(struct f2fs_sb_info *sbi)
2939{
2940 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
2941 struct f2fs_nm_info *nm_i = NM_I(sbi);
2942 unsigned char *version_bitmap;
22ad0b6a
JK
2943 unsigned int nat_segs;
2944 int err;
e05df3b1
JK
2945
2946 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
2947
2948 /* segment_count_nat includes pair segment so divide to 2. */
2949 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
22ad0b6a
JK
2950 nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
2951 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
7ee0eeab 2952
b63da15e 2953 /* not used nids: 0, node, meta, (and root counted as valid node) */
04d47e67 2954 nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count -
292c196a 2955 sbi->nquota_files - F2FS_RESERVED_NODE_NUM;
9a4ffdf5
CY
2956 nm_i->nid_cnt[FREE_NID] = 0;
2957 nm_i->nid_cnt[PREALLOC_NID] = 0;
e05df3b1 2958 nm_i->nat_cnt = 0;
cdfc41c1 2959 nm_i->ram_thresh = DEF_RAM_THRESHOLD;
ea1a29a0 2960 nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
2304cb0c 2961 nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
e05df3b1 2962
8a7ed66a 2963 INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
9a4ffdf5 2964 INIT_LIST_HEAD(&nm_i->free_nid_list);
769ec6e5
JK
2965 INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO);
2966 INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO);
e05df3b1 2967 INIT_LIST_HEAD(&nm_i->nat_entries);
22969158 2968 spin_lock_init(&nm_i->nat_list_lock);
e05df3b1
JK
2969
2970 mutex_init(&nm_i->build_lock);
b8559dc2 2971 spin_lock_init(&nm_i->nid_list_lock);
b873b798 2972 init_rwsem(&nm_i->nat_tree_lock);
e05df3b1 2973
e05df3b1 2974 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
79b5793b 2975 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
e05df3b1
JK
2976 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
2977 if (!version_bitmap)
2978 return -EFAULT;
2979
79b5793b
AG
2980 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
2981 GFP_KERNEL);
2982 if (!nm_i->nat_bitmap)
2983 return -ENOMEM;
599a09b2 2984
22ad0b6a
JK
2985 err = __get_nat_bitmaps(sbi);
2986 if (err)
2987 return err;
2988
599a09b2
CY
2989#ifdef CONFIG_F2FS_CHECK_FS
2990 nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size,
2991 GFP_KERNEL);
2992 if (!nm_i->nat_bitmap_mir)
2993 return -ENOMEM;
2994#endif
2995
e05df3b1
JK
2996 return 0;
2997}
2998
9f7e4a2c 2999static int init_free_nid_cache(struct f2fs_sb_info *sbi)
4ac91242
CY
3000{
3001 struct f2fs_nm_info *nm_i = NM_I(sbi);
bb1105e4 3002 int i;
4ac91242 3003
026f0507
KC
3004 nm_i->free_nid_bitmap =
3005 f2fs_kzalloc(sbi, array_size(sizeof(unsigned char *),
3006 nm_i->nat_blocks),
3007 GFP_KERNEL);
4ac91242
CY
3008 if (!nm_i->free_nid_bitmap)
3009 return -ENOMEM;
3010
bb1105e4
JK
3011 for (i = 0; i < nm_i->nat_blocks; i++) {
3012 nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi,
e15d54d5 3013 f2fs_bitmap_size(NAT_ENTRY_PER_BLOCK), GFP_KERNEL);
68c43a23 3014 if (!nm_i->free_nid_bitmap[i])
bb1105e4
JK
3015 return -ENOMEM;
3016 }
3017
628b3d14 3018 nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8,
4ac91242
CY
3019 GFP_KERNEL);
3020 if (!nm_i->nat_block_bitmap)
3021 return -ENOMEM;
586d1492 3022
9d2a789c
KC
3023 nm_i->free_nid_count =
3024 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned short),
3025 nm_i->nat_blocks),
3026 GFP_KERNEL);
586d1492
CY
3027 if (!nm_i->free_nid_count)
3028 return -ENOMEM;
4ac91242
CY
3029 return 0;
3030}
3031
4d57b86d 3032int f2fs_build_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK
3033{
3034 int err;
3035
acbf054d
CY
3036 sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info),
3037 GFP_KERNEL);
e05df3b1
JK
3038 if (!sbi->nm_info)
3039 return -ENOMEM;
3040
3041 err = init_node_manager(sbi);
3042 if (err)
3043 return err;
3044
4ac91242
CY
3045 err = init_free_nid_cache(sbi);
3046 if (err)
3047 return err;
3048
7041d5d2
CY
3049 /* load free nid status from nat_bits table */
3050 load_free_nid_bitmap(sbi);
3051
e2374015 3052 return f2fs_build_free_nids(sbi, true, true);
e05df3b1
JK
3053}
3054
4d57b86d 3055void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK
3056{
3057 struct f2fs_nm_info *nm_i = NM_I(sbi);
3058 struct free_nid *i, *next_i;
3059 struct nat_entry *natvec[NATVEC_SIZE];
7aed0d45 3060 struct nat_entry_set *setvec[SETVEC_SIZE];
e05df3b1
JK
3061 nid_t nid = 0;
3062 unsigned int found;
3063
3064 if (!nm_i)
3065 return;
3066
3067 /* destroy free nid list */
b8559dc2 3068 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 3069 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
a0761f63 3070 __remove_free_nid(sbi, i, FREE_NID);
b8559dc2 3071 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0 3072 kmem_cache_free(free_nid_slab, i);
b8559dc2 3073 spin_lock(&nm_i->nid_list_lock);
e05df3b1 3074 }
9a4ffdf5
CY
3075 f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]);
3076 f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]);
3077 f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list));
b8559dc2 3078 spin_unlock(&nm_i->nid_list_lock);
e05df3b1
JK
3079
3080 /* destroy nat cache */
b873b798 3081 down_write(&nm_i->nat_tree_lock);
e05df3b1
JK
3082 while ((found = __gang_lookup_nat_cache(nm_i,
3083 nid, NATVEC_SIZE, natvec))) {
3084 unsigned idx;
7aed0d45 3085
b6ce391e 3086 nid = nat_get_nid(natvec[found - 1]) + 1;
22969158
CY
3087 for (idx = 0; idx < found; idx++) {
3088 spin_lock(&nm_i->nat_list_lock);
3089 list_del(&natvec[idx]->list);
3090 spin_unlock(&nm_i->nat_list_lock);
3091
b6ce391e 3092 __del_from_nat_cache(nm_i, natvec[idx]);
22969158 3093 }
e05df3b1 3094 }
9850cf4a 3095 f2fs_bug_on(sbi, nm_i->nat_cnt);
7aed0d45
JK
3096
3097 /* destroy nat set cache */
3098 nid = 0;
3099 while ((found = __gang_lookup_nat_set(nm_i,
3100 nid, SETVEC_SIZE, setvec))) {
3101 unsigned idx;
3102
3103 nid = setvec[found - 1]->set + 1;
3104 for (idx = 0; idx < found; idx++) {
3105 /* entry_cnt is not zero, when cp_error was occurred */
3106 f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list));
3107 radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set);
3108 kmem_cache_free(nat_entry_set_slab, setvec[idx]);
3109 }
3110 }
b873b798 3111 up_write(&nm_i->nat_tree_lock);
e05df3b1 3112
4ac91242 3113 kvfree(nm_i->nat_block_bitmap);
bb1105e4
JK
3114 if (nm_i->free_nid_bitmap) {
3115 int i;
3116
3117 for (i = 0; i < nm_i->nat_blocks; i++)
3118 kvfree(nm_i->free_nid_bitmap[i]);
5222595d 3119 kvfree(nm_i->free_nid_bitmap);
bb1105e4 3120 }
586d1492 3121 kvfree(nm_i->free_nid_count);
4ac91242 3122
5222595d
JK
3123 kvfree(nm_i->nat_bitmap);
3124 kvfree(nm_i->nat_bits);
599a09b2 3125#ifdef CONFIG_F2FS_CHECK_FS
5222595d 3126 kvfree(nm_i->nat_bitmap_mir);
599a09b2 3127#endif
e05df3b1 3128 sbi->nm_info = NULL;
5222595d 3129 kvfree(nm_i);
e05df3b1
JK
3130}
3131
4d57b86d 3132int __init f2fs_create_node_manager_caches(void)
e05df3b1
JK
3133{
3134 nat_entry_slab = f2fs_kmem_cache_create("nat_entry",
e8512d2e 3135 sizeof(struct nat_entry));
e05df3b1 3136 if (!nat_entry_slab)
aec71382 3137 goto fail;
e05df3b1
JK
3138
3139 free_nid_slab = f2fs_kmem_cache_create("free_nid",
e8512d2e 3140 sizeof(struct free_nid));
aec71382 3141 if (!free_nid_slab)
ce3e6d25 3142 goto destroy_nat_entry;
aec71382
CY
3143
3144 nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set",
3145 sizeof(struct nat_entry_set));
3146 if (!nat_entry_set_slab)
ce3e6d25 3147 goto destroy_free_nid;
50fa53ec
CY
3148
3149 fsync_node_entry_slab = f2fs_kmem_cache_create("fsync_node_entry",
3150 sizeof(struct fsync_node_entry));
3151 if (!fsync_node_entry_slab)
3152 goto destroy_nat_entry_set;
e05df3b1 3153 return 0;
aec71382 3154
50fa53ec
CY
3155destroy_nat_entry_set:
3156 kmem_cache_destroy(nat_entry_set_slab);
ce3e6d25 3157destroy_free_nid:
aec71382 3158 kmem_cache_destroy(free_nid_slab);
ce3e6d25 3159destroy_nat_entry:
aec71382
CY
3160 kmem_cache_destroy(nat_entry_slab);
3161fail:
3162 return -ENOMEM;
e05df3b1
JK
3163}
3164
4d57b86d 3165void f2fs_destroy_node_manager_caches(void)
e05df3b1 3166{
50fa53ec 3167 kmem_cache_destroy(fsync_node_entry_slab);
aec71382 3168 kmem_cache_destroy(nat_entry_set_slab);
e05df3b1
JK
3169 kmem_cache_destroy(free_nid_slab);
3170 kmem_cache_destroy(nat_entry_slab);
3171}