projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'integrity-v6.7' of git://git.kernel.org/pub/scm/linux/kernel/git/zohar...
[linux-block.git] / fs / f2fs / node.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
e05df3b1
JK		 3 * fs/f2fs/node.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
e05df3b1
JK		 7 */
8#include <linux/fs.h>
9#include <linux/f2fs_fs.h>
10#include <linux/mpage.h>
4034247a 11#include <linux/sched/mm.h>
e05df3b1
JK		 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"
52118743 20#include "iostat.h"
51dd6249 21#include <trace/events/f2fs.h>
e05df3b1 22
4d57b86d 23#define on_f2fs_build_free_nids(nmi) mutex_is_locked(&(nm_i)->build_lock)
f978f5a0 24
e05df3b1
JK		 25static struct kmem_cache *nat_entry_slab;
26static struct kmem_cache *free_nid_slab;
aec71382 27static struct kmem_cache *nat_entry_set_slab;
50fa53ec 28static struct kmem_cache *fsync_node_entry_slab;
e05df3b1 29
a4f843bd
JK		 30/*
31 * Check whether the given nid is within node id range.
32 */
4d57b86d 33int f2fs_check_nid_range(struct f2fs_sb_info *sbi, nid_t nid)
a4f843bd
JK		 34{
35 if (unlikely(nid < F2FS_ROOT_INO(sbi) || nid >= NM_I(sbi)->max_nid)) {
36 set_sbi_flag(sbi, SBI_NEED_FSCK);
dcbb4c10
JP		 37 f2fs_warn(sbi, "%s: out-of-range nid=%x, run fsck to fix.",
38 __func__, nid);
95fa90c9 39 f2fs_handle_error(sbi, ERROR_CORRUPTED_INODE);
10f966bb 40 return -EFSCORRUPTED;
a4f843bd
JK		 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);
d6d2b491 48 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
cdfc41c1 49 struct sysinfo val;
e5e7ea3c 50 unsigned long avail_ram;
cdfc41c1 51 unsigned long mem_size = 0;
6fb03f3a 52 bool res = false;
cdfc41c1 53
d6d2b491
ST		 54 if (!nm_i)
55 return true;
56
cdfc41c1 57 si_meminfo(&val);
e5e7ea3c
JK		 58
59 /* only uses low memory */
60 avail_ram = val.totalram - val.totalhigh;
61
429511cd 62 /*
71644dff 63 * give 25%, 25%, 50%, 50%, 25%, 25% memory for each components respectively
429511cd 64 */
6fb03f3a 65 if (type == FREE_NIDS) {
9a4ffdf5 66 mem_size = (nm_i->nid_cnt[FREE_NID] *
b8559dc2 67 sizeof(struct free_nid)) >> PAGE_SHIFT;
e5e7ea3c 68 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
6fb03f3a 69 } else if (type == NAT_ENTRIES) {
a95ba66a
JK		 70 mem_size = (nm_i->nat_cnt[TOTAL_NAT] *
71 sizeof(struct nat_entry)) >> PAGE_SHIFT;
e5e7ea3c 72 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
e589c2c4
JK		 73 if (excess_cached_nats(sbi))
74 res = false;
a1257023
JK		 75 } else if (type == DIRTY_DENTS) {
76 if (sbi->sb->s_bdi->wb.dirty_exceeded)
77 return false;
78 mem_size = get_pages(sbi, F2FS_DIRTY_DENTS);
79 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
e5e7ea3c
JK		 80 } else if (type == INO_ENTRIES) {
81 int i;
82
39d787be 83 for (i = 0; i < MAX_INO_ENTRY; i++)
8f73cbb7
KM		 84 mem_size += sbi->im[i].ino_num *
85 sizeof(struct ino_entry);
86 mem_size >>= PAGE_SHIFT;
e5e7ea3c 87 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 1);
71644dff
JK		 88 } else if (type == READ_EXTENT_CACHE || type == AGE_EXTENT_CACHE) {
89 enum extent_type etype = type == READ_EXTENT_CACHE ?
90 EX_READ : EX_BLOCK_AGE;
91 struct extent_tree_info *eti = &sbi->extent_tree[etype];
e7547dac
JK		 92
93 mem_size = (atomic_read(&eti->total_ext_tree) *
7441ccef 94 sizeof(struct extent_tree) +
e7547dac 95 atomic_read(&eti->total_ext_node) *
09cbfeaf 96 sizeof(struct extent_node)) >> PAGE_SHIFT;
71644dff 97 res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
d6d2b491
ST		 98 } else if (type == DISCARD_CACHE) {
99 mem_size = (atomic_read(&dcc->discard_cmd_cnt) *
100 sizeof(struct discard_cmd)) >> PAGE_SHIFT;
101 res = mem_size < (avail_ram * nm_i->ram_thresh / 100);
6ce19aff
CY		 102 } else if (type == COMPRESS_PAGE) {
103#ifdef CONFIG_F2FS_FS_COMPRESSION
104 unsigned long free_ram = val.freeram;
105
106 /*
107 * free memory is lower than watermark or cached page count
108 * exceed threshold, deny caching compress page.
109 */
110 res = (free_ram > avail_ram * sbi->compress_watermark / 100) &&
111 (COMPRESS_MAPPING(sbi)->nrpages <
112 free_ram * sbi->compress_percent / 100);
113#else
114 res = false;
115#endif
1e84371f 116 } else {
1663cae4
JK		 117 if (!sbi->sb->s_bdi->wb.dirty_exceeded)
118 return true;
6fb03f3a
JK		 119 }
120 return res;
cdfc41c1
JK		 121}
122
e05df3b1
JK		 123static void clear_node_page_dirty(struct page *page)
124{
e05df3b1 125 if (PageDirty(page)) {
5ec2d99d 126 f2fs_clear_page_cache_dirty_tag(page);
e05df3b1 127 clear_page_dirty_for_io(page);
aec2f729 128 dec_page_count(F2FS_P_SB(page), F2FS_DIRTY_NODES);
e05df3b1
JK		 129 }
130 ClearPageUptodate(page);
131}
132
133static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
134{
3acc4522 135 return f2fs_get_meta_page_retry(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK		 136}
137
138static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
139{
140 struct page *src_page;
141 struct page *dst_page;
e05df3b1
JK		 142 pgoff_t dst_off;
143 void *src_addr;
144 void *dst_addr;
145 struct f2fs_nm_info *nm_i = NM_I(sbi);
146
80551d17 147 dst_off = next_nat_addr(sbi, current_nat_addr(sbi, nid));
e05df3b1
JK		 148
149 /* get current nat block page with lock */
80551d17 150 src_page = get_current_nat_page(sbi, nid);
edc55aaf
JK		 151 if (IS_ERR(src_page))
152 return src_page;
4d57b86d 153 dst_page = f2fs_grab_meta_page(sbi, dst_off);
9850cf4a 154 f2fs_bug_on(sbi, PageDirty(src_page));
e05df3b1
JK		 155
156 src_addr = page_address(src_page);
157 dst_addr = page_address(dst_page);
09cbfeaf 158 memcpy(dst_addr, src_addr, PAGE_SIZE);
e05df3b1
JK		 159 set_page_dirty(dst_page);
160 f2fs_put_page(src_page, 1);
161
162 set_to_next_nat(nm_i, nid);
163
164 return dst_page;
165}
166
32410577
CY		 167static struct nat_entry *__alloc_nat_entry(struct f2fs_sb_info *sbi,
168 nid_t nid, bool no_fail)
12f9ef37
YH		 169{
170 struct nat_entry *new;
171
32410577
CY		 172 new = f2fs_kmem_cache_alloc(nat_entry_slab,
173 GFP_F2FS_ZERO, no_fail, sbi);
12f9ef37
YH		 174 if (new) {
175 nat_set_nid(new, nid);
176 nat_reset_flag(new);
177 }
178 return new;
179}
180
181static void __free_nat_entry(struct nat_entry *e)
182{
183 kmem_cache_free(nat_entry_slab, e);
184}
185
186/* must be locked by nat_tree_lock */
187static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
188 struct nat_entry *ne, struct f2fs_nat_entry *raw_ne, bool no_fail)
189{
190 if (no_fail)
191 f2fs_radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne);
192 else if (radix_tree_insert(&nm_i->nat_root, nat_get_nid(ne), ne))
193 return NULL;
194
195 if (raw_ne)
196 node_info_from_raw_nat(&ne->ni, raw_ne);
22969158
CY		 197
198 spin_lock(&nm_i->nat_list_lock);
12f9ef37 199 list_add_tail(&ne->list, &nm_i->nat_entries);
22969158
CY		 200 spin_unlock(&nm_i->nat_list_lock);
201
a95ba66a
JK		 202 nm_i->nat_cnt[TOTAL_NAT]++;
203 nm_i->nat_cnt[RECLAIMABLE_NAT]++;
12f9ef37
YH		 204 return ne;
205}
206
e05df3b1
JK		 207static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n)
208{
22969158
CY		 209 struct nat_entry *ne;
210
211 ne = radix_tree_lookup(&nm_i->nat_root, n);
212
213 /* for recent accessed nat entry, move it to tail of lru list */
214 if (ne && !get_nat_flag(ne, IS_DIRTY)) {
215 spin_lock(&nm_i->nat_list_lock);
216 if (!list_empty(&ne->list))
217 list_move_tail(&ne->list, &nm_i->nat_entries);
218 spin_unlock(&nm_i->nat_list_lock);
219 }
220
221 return ne;
e05df3b1
JK		 222}
223
224static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
225 nid_t start, unsigned int nr, struct nat_entry **ep)
226{
227 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr);
228}
229
230static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
231{
e05df3b1 232 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
a95ba66a
JK		 233 nm_i->nat_cnt[TOTAL_NAT]--;
234 nm_i->nat_cnt[RECLAIMABLE_NAT]--;
12f9ef37 235 __free_nat_entry(e);
e05df3b1
JK		 236}
237
780de47c
CY		 238static struct nat_entry_set *__grab_nat_entry_set(struct f2fs_nm_info *nm_i,
239 struct nat_entry *ne)
309cc2b6
JK		 240{
241 nid_t set = NAT_BLOCK_OFFSET(ne->ni.nid);
242 struct nat_entry_set *head;
243
309cc2b6
JK		 244 head = radix_tree_lookup(&nm_i->nat_set_root, set);
245 if (!head) {
32410577
CY		 246 head = f2fs_kmem_cache_alloc(nat_entry_set_slab,
247 GFP_NOFS, true, NULL);
309cc2b6
JK		 248
249 INIT_LIST_HEAD(&head->entry_list);
250 INIT_LIST_HEAD(&head->set_list);
251 head->set = set;
252 head->entry_cnt = 0;
9be32d72 253 f2fs_radix_tree_insert(&nm_i->nat_set_root, set, head);
309cc2b6 254 }
780de47c
CY		 255 return head;
256}
257
258static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
259 struct nat_entry *ne)
260{
261 struct nat_entry_set *head;
262 bool new_ne = nat_get_blkaddr(ne) == NEW_ADDR;
263
264 if (!new_ne)
265 head = __grab_nat_entry_set(nm_i, ne);
266
267 /*
268 * update entry_cnt in below condition:
269 * 1. update NEW_ADDR to valid block address;
270 * 2. update old block address to new one;
271 */
272 if (!new_ne && (get_nat_flag(ne, IS_PREALLOC) ||
273 !get_nat_flag(ne, IS_DIRTY)))
274 head->entry_cnt++;
275
276 set_nat_flag(ne, IS_PREALLOC, new_ne);
febeca6d
CY		 277
278 if (get_nat_flag(ne, IS_DIRTY))
279 goto refresh_list;
280
a95ba66a
JK		 281 nm_i->nat_cnt[DIRTY_NAT]++;
282 nm_i->nat_cnt[RECLAIMABLE_NAT]--;
309cc2b6 283 set_nat_flag(ne, IS_DIRTY, true);
febeca6d 284refresh_list:
22969158 285 spin_lock(&nm_i->nat_list_lock);
780de47c 286 if (new_ne)
febeca6d
CY		 287 list_del_init(&ne->list);
288 else
289 list_move_tail(&ne->list, &head->entry_list);
22969158 290 spin_unlock(&nm_i->nat_list_lock);
309cc2b6
JK		 291}
292
293static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
0b28b71e 294 struct nat_entry_set *set, struct nat_entry *ne)
309cc2b6 295{
22969158 296 spin_lock(&nm_i->nat_list_lock);
0b28b71e 297 list_move_tail(&ne->list, &nm_i->nat_entries);
22969158
CY		 298 spin_unlock(&nm_i->nat_list_lock);
299
0b28b71e
KM		 300 set_nat_flag(ne, IS_DIRTY, false);
301 set->entry_cnt--;
a95ba66a
JK		 302 nm_i->nat_cnt[DIRTY_NAT]--;
303 nm_i->nat_cnt[RECLAIMABLE_NAT]++;
309cc2b6
JK		 304}
305
306static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
307 nid_t start, unsigned int nr, struct nat_entry_set **ep)
308{
309 return radix_tree_gang_lookup(&nm_i->nat_set_root, (void **)ep,
310 start, nr);
311}
312
50fa53ec
CY		 313bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page)
314{
315 return NODE_MAPPING(sbi) == page->mapping &&
316 IS_DNODE(page) && is_cold_node(page);
317}
318
319void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi)
320{
321 spin_lock_init(&sbi->fsync_node_lock);
322 INIT_LIST_HEAD(&sbi->fsync_node_list);
323 sbi->fsync_seg_id = 0;
324 sbi->fsync_node_num = 0;
325}
326
327static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi,
328 struct page *page)
329{
330 struct fsync_node_entry *fn;
331 unsigned long flags;
332 unsigned int seq_id;
333
32410577
CY		 334 fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab,
335 GFP_NOFS, true, NULL);
50fa53ec
CY		 336
337 get_page(page);
338 fn->page = page;
339 INIT_LIST_HEAD(&fn->list);
340
341 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
342 list_add_tail(&fn->list, &sbi->fsync_node_list);
343 fn->seq_id = sbi->fsync_seg_id++;
344 seq_id = fn->seq_id;
345 sbi->fsync_node_num++;
346 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
347
348 return seq_id;
349}
350
351void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page)
352{
353 struct fsync_node_entry *fn;
354 unsigned long flags;
355
356 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
357 list_for_each_entry(fn, &sbi->fsync_node_list, list) {
358 if (fn->page == page) {
359 list_del(&fn->list);
360 sbi->fsync_node_num--;
361 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
362 kmem_cache_free(fsync_node_entry_slab, fn);
363 put_page(page);
364 return;
365 }
366 }
367 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
368 f2fs_bug_on(sbi, 1);
369}
370
371void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi)
372{
373 unsigned long flags;
374
375 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
376 sbi->fsync_seg_id = 0;
377 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
378}
379
4d57b86d 380int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK		 381{
382 struct f2fs_nm_info *nm_i = NM_I(sbi);
383 struct nat_entry *e;
2dcf51ab 384 bool need = false;
e05df3b1 385
e4544b63 386 f2fs_down_read(&nm_i->nat_tree_lock);
e05df3b1 387 e = __lookup_nat_cache(nm_i, nid);
2dcf51ab
JK		 388 if (e) {
389 if (!get_nat_flag(e, IS_CHECKPOINTED) &&
390 !get_nat_flag(e, HAS_FSYNCED_INODE))
391 need = true;
392 }
e4544b63 393 f2fs_up_read(&nm_i->nat_tree_lock);
2dcf51ab 394 return need;
e05df3b1
JK		 395}
396
4d57b86d 397bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid)
479f40c4
JK		 398{
399 struct f2fs_nm_info *nm_i = NM_I(sbi);
400 struct nat_entry *e;
2dcf51ab 401 bool is_cp = true;
479f40c4 402
e4544b63 403 f2fs_down_read(&nm_i->nat_tree_lock);
2dcf51ab
JK		 404 e = __lookup_nat_cache(nm_i, nid);
405 if (e && !get_nat_flag(e, IS_CHECKPOINTED))
406 is_cp = false;
e4544b63 407 f2fs_up_read(&nm_i->nat_tree_lock);
2dcf51ab 408 return is_cp;
479f40c4
JK		 409}
410
4d57b86d 411bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino)
b6fe5873
JK		 412{
413 struct f2fs_nm_info *nm_i = NM_I(sbi);
414 struct nat_entry *e;
88bd02c9 415 bool need_update = true;
b6fe5873 416
e4544b63 417 f2fs_down_read(&nm_i->nat_tree_lock);
88bd02c9
JK		 418 e = __lookup_nat_cache(nm_i, ino);
419 if (e && get_nat_flag(e, HAS_LAST_FSYNC) &&
420 (get_nat_flag(e, IS_CHECKPOINTED) ||
421 get_nat_flag(e, HAS_FSYNCED_INODE)))
422 need_update = false;
e4544b63 423 f2fs_up_read(&nm_i->nat_tree_lock);
88bd02c9 424 return need_update;
b6fe5873
JK		 425}
426
12f9ef37 427/* must be locked by nat_tree_lock */
1515aef0 428static void cache_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
e05df3b1
JK		 429 struct f2fs_nat_entry *ne)
430{
1515aef0 431 struct f2fs_nm_info *nm_i = NM_I(sbi);
12f9ef37 432 struct nat_entry *new, *e;
9be32d72 433
0df035c7 434 /* Let's mitigate lock contention of nat_tree_lock during checkpoint */
e4544b63 435 if (f2fs_rwsem_is_locked(&sbi->cp_global_sem))
0df035c7
JK		 436 return;
437
32410577 438 new = __alloc_nat_entry(sbi, nid, false);
12f9ef37
YH		 439 if (!new)
440 return;
441
e4544b63 442 f2fs_down_write(&nm_i->nat_tree_lock);
e05df3b1 443 e = __lookup_nat_cache(nm_i, nid);
12f9ef37
YH		 444 if (!e)
445 e = __init_nat_entry(nm_i, new, ne, false);
446 else
0c0b471e
EB		 447 f2fs_bug_on(sbi, nat_get_ino(e) != le32_to_cpu(ne->ino) ||
448 nat_get_blkaddr(e) !=
449 le32_to_cpu(ne->block_addr) ||
1515aef0 450 nat_get_version(e) != ne->version);
e4544b63 451 f2fs_up_write(&nm_i->nat_tree_lock);
12f9ef37
YH		 452 if (e != new)
453 __free_nat_entry(new);
e05df3b1
JK		 454}
455
456static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni,
479f40c4 457 block_t new_blkaddr, bool fsync_done)
e05df3b1
JK		 458{
459 struct f2fs_nm_info *nm_i = NM_I(sbi);
460 struct nat_entry *e;
32410577 461 struct nat_entry *new = __alloc_nat_entry(sbi, ni->nid, true);
9be32d72 462
e4544b63 463 f2fs_down_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 464 e = __lookup_nat_cache(nm_i, ni->nid);
465 if (!e) {
12f9ef37 466 e = __init_nat_entry(nm_i, new, NULL, true);
5c27f4ee 467 copy_node_info(&e->ni, ni);
9850cf4a 468 f2fs_bug_on(sbi, ni->blk_addr == NEW_ADDR);
e05df3b1
JK		 469 } else if (new_blkaddr == NEW_ADDR) {
470 /*
471 * when nid is reallocated,
472 * previous nat entry can be remained in nat cache.
473 * So, reinitialize it with new information.
474 */
5c27f4ee 475 copy_node_info(&e->ni, ni);
9850cf4a 476 f2fs_bug_on(sbi, ni->blk_addr != NULL_ADDR);
e05df3b1 477 }
12f9ef37
YH		 478 /* let's free early to reduce memory consumption */
479 if (e != new)
480 __free_nat_entry(new);
e05df3b1 481
e05df3b1 482 /* sanity check */
9850cf4a
JK		 483 f2fs_bug_on(sbi, nat_get_blkaddr(e) != ni->blk_addr);
484 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NULL_ADDR &&
e05df3b1 485 new_blkaddr == NULL_ADDR);
9850cf4a 486 f2fs_bug_on(sbi, nat_get_blkaddr(e) == NEW_ADDR &&
e05df3b1 487 new_blkaddr == NEW_ADDR);
93770ab7 488 f2fs_bug_on(sbi, __is_valid_data_blkaddr(nat_get_blkaddr(e)) &&
e05df3b1
JK		 489 new_blkaddr == NEW_ADDR);
490
e1c42045 491 /* increment version no as node is removed */
e05df3b1
JK		 492 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) {
493 unsigned char version = nat_get_version(e);
5f029c04 494
e05df3b1
JK		 495 nat_set_version(e, inc_node_version(version));
496 }
497
498 /* change address */
499 nat_set_blkaddr(e, new_blkaddr);
93770ab7 500 if (!__is_valid_data_blkaddr(new_blkaddr))
88bd02c9 501 set_nat_flag(e, IS_CHECKPOINTED, false);
e05df3b1 502 __set_nat_cache_dirty(nm_i, e);
479f40c4
JK		 503
504 /* update fsync_mark if its inode nat entry is still alive */
d5b692b7
CY		 505 if (ni->nid != ni->ino)
506 e = __lookup_nat_cache(nm_i, ni->ino);
88bd02c9
JK		 507 if (e) {
508 if (fsync_done && ni->nid == ni->ino)
509 set_nat_flag(e, HAS_FSYNCED_INODE, true);
510 set_nat_flag(e, HAS_LAST_FSYNC, fsync_done);
511 }
e4544b63 512 f2fs_up_write(&nm_i->nat_tree_lock);
e05df3b1
JK		 513}
514
4d57b86d 515int f2fs_try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink)
e05df3b1
JK		 516{
517 struct f2fs_nm_info *nm_i = NM_I(sbi);
1b38dc8e 518 int nr = nr_shrink;
e05df3b1 519
e4544b63 520 if (!f2fs_down_write_trylock(&nm_i->nat_tree_lock))
b873b798 521 return 0;
e05df3b1 522
22969158
CY		 523 spin_lock(&nm_i->nat_list_lock);
524 while (nr_shrink) {
e05df3b1 525 struct nat_entry *ne;
22969158
CY		 526
527 if (list_empty(&nm_i->nat_entries))
528 break;
529
e05df3b1
JK		 530 ne = list_first_entry(&nm_i->nat_entries,
531 struct nat_entry, list);
22969158
CY		 532 list_del(&ne->list);
533 spin_unlock(&nm_i->nat_list_lock);
534
e05df3b1
JK		 535 __del_from_nat_cache(nm_i, ne);
536 nr_shrink--;
22969158
CY		 537
538 spin_lock(&nm_i->nat_list_lock);
e05df3b1 539 }
22969158
CY		 540 spin_unlock(&nm_i->nat_list_lock);
541
e4544b63 542 f2fs_up_write(&nm_i->nat_tree_lock);
1b38dc8e 543 return nr - nr_shrink;
e05df3b1
JK		 544}
545
7735730d 546int f2fs_get_node_info(struct f2fs_sb_info *sbi, nid_t nid,
a9419b63 547 struct node_info *ni, bool checkpoint_context)
e05df3b1
JK		 548{
549 struct f2fs_nm_info *nm_i = NM_I(sbi);
550 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 551 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 552 nid_t start_nid = START_NID(nid);
553 struct f2fs_nat_block *nat_blk;
554 struct page *page = NULL;
555 struct f2fs_nat_entry ne;
556 struct nat_entry *e;
66a82d1f 557 pgoff_t index;
93770ab7 558 block_t blkaddr;
e05df3b1
JK		 559 int i;
560
561 ni->nid = nid;
2eeb0dce 562retry:
e05df3b1 563 /* Check nat cache */
e4544b63 564 f2fs_down_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 565 e = __lookup_nat_cache(nm_i, nid);
566 if (e) {
567 ni->ino = nat_get_ino(e);
568 ni->blk_addr = nat_get_blkaddr(e);
569 ni->version = nat_get_version(e);
e4544b63 570 f2fs_up_read(&nm_i->nat_tree_lock);
7735730d 571 return 0;
1515aef0 572 }
e05df3b1 573
2eeb0dce
JK		 574 /*
575 * Check current segment summary by trying to grab journal_rwsem first.
576 * This sem is on the critical path on the checkpoint requiring the above
577 * nat_tree_lock. Therefore, we should retry, if we failed to grab here
578 * while not bothering checkpoint.
579 */
e4544b63 580 if (!f2fs_rwsem_is_locked(&sbi->cp_global_sem) || checkpoint_context) {
2eeb0dce 581 down_read(&curseg->journal_rwsem);
e4544b63 582 } else if (f2fs_rwsem_is_contended(&nm_i->nat_tree_lock) ||
a9419b63 583 !down_read_trylock(&curseg->journal_rwsem)) {
e4544b63 584 f2fs_up_read(&nm_i->nat_tree_lock);
2eeb0dce
JK		 585 goto retry;
586 }
3547ea96 587
4d57b86d 588 i = f2fs_lookup_journal_in_cursum(journal, NAT_JOURNAL, nid, 0);
e05df3b1 589 if (i >= 0) {
dfc08a12 590 ne = nat_in_journal(journal, i);
e05df3b1
JK		 591 node_info_from_raw_nat(ni, &ne);
592 }
544b53da 593 up_read(&curseg->journal_rwsem);
66a82d1f 594 if (i >= 0) {
e4544b63 595 f2fs_up_read(&nm_i->nat_tree_lock);
e05df3b1 596 goto cache;
66a82d1f 597 }
e05df3b1
JK		 598
599 /* Fill node_info from nat page */
66a82d1f 600 index = current_nat_addr(sbi, nid);
e4544b63 601 f2fs_up_read(&nm_i->nat_tree_lock);
66a82d1f 602
4d57b86d 603 page = f2fs_get_meta_page(sbi, index);
7735730d
CY		 604 if (IS_ERR(page))
605 return PTR_ERR(page);
606
e05df3b1
JK		 607 nat_blk = (struct f2fs_nat_block *)page_address(page);
608 ne = nat_blk->entries[nid - start_nid];
609 node_info_from_raw_nat(ni, &ne);
610 f2fs_put_page(page, 1);
611cache:
93770ab7
CY		 612 blkaddr = le32_to_cpu(ne.block_addr);
613 if (__is_valid_data_blkaddr(blkaddr) &&
614 !f2fs_is_valid_blkaddr(sbi, blkaddr, DATA_GENERIC_ENHANCE))
615 return -EFAULT;
616
e05df3b1 617 /* cache nat entry */
1515aef0 618 cache_nat_entry(sbi, nid, &ne);
7735730d 619 return 0;
e05df3b1
JK		 620}
621
79344efb
JK		 622/*
623 * readahead MAX_RA_NODE number of node pages.
624 */
4d57b86d 625static void f2fs_ra_node_pages(struct page *parent, int start, int n)
79344efb
JK		 626{
627 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
628 struct blk_plug plug;
629 int i, end;
630 nid_t nid;
631
632 blk_start_plug(&plug);
633
634 /* Then, try readahead for siblings of the desired node */
635 end = start + n;
636 end = min(end, NIDS_PER_BLOCK);
637 for (i = start; i < end; i++) {
638 nid = get_nid(parent, i, false);
4d57b86d 639 f2fs_ra_node_page(sbi, nid);
79344efb
JK		 640 }
641
642 blk_finish_plug(&plug);
643}
644
4d57b86d 645pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs)
3cf45747
CY		 646{
647 const long direct_index = ADDRS_PER_INODE(dn->inode);
d02a6e61
CY		 648 const long direct_blks = ADDRS_PER_BLOCK(dn->inode);
649 const long indirect_blks = ADDRS_PER_BLOCK(dn->inode) * NIDS_PER_BLOCK;
650 unsigned int skipped_unit = ADDRS_PER_BLOCK(dn->inode);
3cf45747
CY		 651 int cur_level = dn->cur_level;
652 int max_level = dn->max_level;
653 pgoff_t base = 0;
654
655 if (!dn->max_level)
656 return pgofs + 1;
657
658 while (max_level-- > cur_level)
659 skipped_unit *= NIDS_PER_BLOCK;
660
661 switch (dn->max_level) {
662 case 3:
663 base += 2 * indirect_blks;
df561f66 664 fallthrough;
3cf45747
CY		 665 case 2:
666 base += 2 * direct_blks;
df561f66 667 fallthrough;
3cf45747
CY		 668 case 1:
669 base += direct_index;
670 break;
671 default:
672 f2fs_bug_on(F2FS_I_SB(dn->inode), 1);
673 }
674
675 return ((pgofs - base) / skipped_unit + 1) * skipped_unit + base;
676}
677
0a8165d7 678/*
e05df3b1
JK		 679 * The maximum depth is four.
680 * Offset[0] will have raw inode offset.
681 */
81ca7350 682static int get_node_path(struct inode *inode, long block,
de93653f 683 int offset[4], unsigned int noffset[4])
e05df3b1 684{
81ca7350 685 const long direct_index = ADDRS_PER_INODE(inode);
d02a6e61 686 const long direct_blks = ADDRS_PER_BLOCK(inode);
e05df3b1 687 const long dptrs_per_blk = NIDS_PER_BLOCK;
d02a6e61 688 const long indirect_blks = ADDRS_PER_BLOCK(inode) * NIDS_PER_BLOCK;
e05df3b1
JK		 689 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK;
690 int n = 0;
691 int level = 0;
692
693 noffset[0] = 0;
694
695 if (block < direct_index) {
25c0a6e5 696 offset[n] = block;
e05df3b1
JK		 697 goto got;
698 }
699 block -= direct_index;
700 if (block < direct_blks) {
701 offset[n++] = NODE_DIR1_BLOCK;
702 noffset[n] = 1;
25c0a6e5 703 offset[n] = block;
e05df3b1
JK		 704 level = 1;
705 goto got;
706 }
707 block -= direct_blks;
708 if (block < direct_blks) {
709 offset[n++] = NODE_DIR2_BLOCK;
710 noffset[n] = 2;
25c0a6e5 711 offset[n] = block;
e05df3b1
JK		 712 level = 1;
713 goto got;
714 }
715 block -= direct_blks;
716 if (block < indirect_blks) {
717 offset[n++] = NODE_IND1_BLOCK;
718 noffset[n] = 3;
719 offset[n++] = block / direct_blks;
720 noffset[n] = 4 + offset[n - 1];
25c0a6e5 721 offset[n] = block % direct_blks;
e05df3b1
JK		 722 level = 2;
723 goto got;
724 }
725 block -= indirect_blks;
726 if (block < indirect_blks) {
727 offset[n++] = NODE_IND2_BLOCK;
728 noffset[n] = 4 + dptrs_per_blk;
729 offset[n++] = block / direct_blks;
730 noffset[n] = 5 + dptrs_per_blk + offset[n - 1];
25c0a6e5 731 offset[n] = block % direct_blks;
e05df3b1
JK		 732 level = 2;
733 goto got;
734 }
735 block -= indirect_blks;
736 if (block < dindirect_blks) {
737 offset[n++] = NODE_DIND_BLOCK;
738 noffset[n] = 5 + (dptrs_per_blk * 2);
739 offset[n++] = block / indirect_blks;
740 noffset[n] = 6 + (dptrs_per_blk * 2) +
741 offset[n - 1] * (dptrs_per_blk + 1);
742 offset[n++] = (block / direct_blks) % dptrs_per_blk;
743 noffset[n] = 7 + (dptrs_per_blk * 2) +
744 offset[n - 2] * (dptrs_per_blk + 1) +
745 offset[n - 1];
25c0a6e5 746 offset[n] = block % direct_blks;
e05df3b1
JK		 747 level = 3;
748 goto got;
749 } else {
adb6dc19 750 return -E2BIG;
e05df3b1
JK		 751 }
752got:
753 return level;
754}
755
756/*
757 * Caller should call f2fs_put_dnode(dn).
4f4124d0 758 * Also, it should grab and release a rwsem by calling f2fs_lock_op() and
7a88ddb5 759 * f2fs_unlock_op() only if mode is set with ALLOC_NODE.
e05df3b1 760 */
4d57b86d 761int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
e05df3b1 762{
4081363f 763 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 764 struct page *npage[4];
f1a3b98e 765 struct page *parent = NULL;
e05df3b1
JK		 766 int offset[4];
767 unsigned int noffset[4];
768 nid_t nids[4];
3cf45747 769 int level, i = 0;
e05df3b1
JK		 770 int err = 0;
771
81ca7350 772 level = get_node_path(dn->inode, index, offset, noffset);
adb6dc19
JK		 773 if (level < 0)
774 return level;
e05df3b1
JK		 775
776 nids[0] = dn->inode->i_ino;
1646cfac 777 npage[0] = dn->inode_page;
e05df3b1 778
1646cfac 779 if (!npage[0]) {
4d57b86d 780 npage[0] = f2fs_get_node_page(sbi, nids[0]);
1646cfac
JK		 781 if (IS_ERR(npage[0]))
782 return PTR_ERR(npage[0]);
783 }
f1a3b98e
JK		 784
785 /* if inline_data is set, should not report any block indices */
786 if (f2fs_has_inline_data(dn->inode) && index) {
76629165 787 err = -ENOENT;
f1a3b98e
JK		 788 f2fs_put_page(npage[0], 1);
789 goto release_out;
790 }
791
e05df3b1 792 parent = npage[0];
52c2db3f
CL		 793 if (level != 0)
794 nids[1] = get_nid(parent, offset[0], true);
e05df3b1
JK		 795 dn->inode_page = npage[0];
796 dn->inode_page_locked = true;
797
798 /* get indirect or direct nodes */
799 for (i = 1; i <= level; i++) {
800 bool done = false;
801
266e97a8 802 if (!nids[i] && mode == ALLOC_NODE) {
e05df3b1 803 /* alloc new node */
4d57b86d 804 if (!f2fs_alloc_nid(sbi, &(nids[i]))) {
e05df3b1
JK		 805 err = -ENOSPC;
806 goto release_pages;
807 }
808
809 dn->nid = nids[i];
4d57b86d 810 npage[i] = f2fs_new_node_page(dn, noffset[i]);
e05df3b1 811 if (IS_ERR(npage[i])) {
4d57b86d 812 f2fs_alloc_nid_failed(sbi, nids[i]);
e05df3b1
JK		 813 err = PTR_ERR(npage[i]);
814 goto release_pages;
815 }
816
817 set_nid(parent, offset[i - 1], nids[i], i == 1);
4d57b86d 818 f2fs_alloc_nid_done(sbi, nids[i]);
e05df3b1 819 done = true;
266e97a8 820 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) {
4d57b86d 821 npage[i] = f2fs_get_node_page_ra(parent, offset[i - 1]);
e05df3b1
JK		 822 if (IS_ERR(npage[i])) {
823 err = PTR_ERR(npage[i]);
824 goto release_pages;
825 }
826 done = true;
827 }
828 if (i == 1) {
829 dn->inode_page_locked = false;
830 unlock_page(parent);
831 } else {
832 f2fs_put_page(parent, 1);
833 }
834
835 if (!done) {
4d57b86d 836 npage[i] = f2fs_get_node_page(sbi, nids[i]);
e05df3b1
JK		 837 if (IS_ERR(npage[i])) {
838 err = PTR_ERR(npage[i]);
839 f2fs_put_page(npage[0], 0);
840 goto release_out;
841 }
842 }
843 if (i < level) {
844 parent = npage[i];
845 nids[i + 1] = get_nid(parent, offset[i], false);
846 }
847 }
848 dn->nid = nids[level];
849 dn->ofs_in_node = offset[level];
850 dn->node_page = npage[level];
a2ced1ce 851 dn->data_blkaddr = f2fs_data_blkaddr(dn);
94afd6d6
CY		 852
853 if (is_inode_flag_set(dn->inode, FI_COMPRESSED_FILE) &&
854 f2fs_sb_has_readonly(sbi)) {
855 unsigned int c_len = f2fs_cluster_blocks_are_contiguous(dn);
856 block_t blkaddr;
857
858 if (!c_len)
859 goto out;
860
861 blkaddr = f2fs_data_blkaddr(dn);
862 if (blkaddr == COMPRESS_ADDR)
863 blkaddr = data_blkaddr(dn->inode, dn->node_page,
864 dn->ofs_in_node + 1);
865
e7547dac 866 f2fs_update_read_extent_tree_range_compressed(dn->inode,
94afd6d6
CY		 867 index, blkaddr,
868 F2FS_I(dn->inode)->i_cluster_size,
869 c_len);
870 }
871out:
e05df3b1
JK		 872 return 0;
873
874release_pages:
875 f2fs_put_page(parent, 1);
876 if (i > 1)
877 f2fs_put_page(npage[0], 0);
878release_out:
879 dn->inode_page = NULL;
880 dn->node_page = NULL;
3cf45747
CY		 881 if (err == -ENOENT) {
882 dn->cur_level = i;
883 dn->max_level = level;
0a2aa8fb 884 dn->ofs_in_node = offset[level];
3cf45747 885 }
e05df3b1
JK		 886 return err;
887}
888
7735730d 889static int truncate_node(struct dnode_of_data *dn)
e05df3b1 890{
4081363f 891 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 892 struct node_info ni;
7735730d 893 int err;
0ea295dd 894 pgoff_t index;
e05df3b1 895
a9419b63 896 err = f2fs_get_node_info(sbi, dn->nid, &ni, false);
7735730d
CY		 897 if (err)
898 return err;
e05df3b1 899
e05df3b1 900 /* Deallocate node address */
4d57b86d 901 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 902 dec_valid_node_count(sbi, dn->inode, dn->nid == dn->inode->i_ino);
479f40c4 903 set_node_addr(sbi, &ni, NULL_ADDR, false);
e05df3b1
JK		 904
905 if (dn->nid == dn->inode->i_ino) {
4d57b86d 906 f2fs_remove_orphan_inode(sbi, dn->nid);
e05df3b1 907 dec_valid_inode_count(sbi);
0f18b462 908 f2fs_inode_synced(dn->inode);
e05df3b1 909 }
000519f2 910
e05df3b1 911 clear_node_page_dirty(dn->node_page);
caf0047e 912 set_sbi_flag(sbi, SBI_IS_DIRTY);
e05df3b1 913
0ea295dd 914 index = dn->node_page->index;
e05df3b1 915 f2fs_put_page(dn->node_page, 1);
bf39c00a
JK		 916
917 invalidate_mapping_pages(NODE_MAPPING(sbi),
0ea295dd 918 index, index);
bf39c00a 919
e05df3b1 920 dn->node_page = NULL;
51dd6249 921 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr);
7735730d
CY		 922
923 return 0;
e05df3b1
JK		 924}
925
926static int truncate_dnode(struct dnode_of_data *dn)
927{
a6ec8378 928 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
e05df3b1 929 struct page *page;
7735730d 930 int err;
e05df3b1
JK		 931
932 if (dn->nid == 0)
933 return 1;
934
935 /* get direct node */
a6ec8378 936 page = f2fs_get_node_page(sbi, dn->nid);
45586c70 937 if (PTR_ERR(page) == -ENOENT)
e05df3b1
JK		 938 return 1;
939 else if (IS_ERR(page))
940 return PTR_ERR(page);
941
a6ec8378
CY		 942 if (IS_INODE(page) || ino_of_node(page) != dn->inode->i_ino) {
943 f2fs_err(sbi, "incorrect node reference, ino: %lu, nid: %u, ino_of_node: %u",
944 dn->inode->i_ino, dn->nid, ino_of_node(page));
945 set_sbi_flag(sbi, SBI_NEED_FSCK);
946 f2fs_handle_error(sbi, ERROR_INVALID_NODE_REFERENCE);
947 f2fs_put_page(page, 1);
948 return -EFSCORRUPTED;
949 }
950
e05df3b1
JK		 951 /* Make dnode_of_data for parameter */
952 dn->node_page = page;
953 dn->ofs_in_node = 0;
a6ec8378 954 f2fs_truncate_data_blocks_range(dn, ADDRS_PER_BLOCK(dn->inode));
7735730d 955 err = truncate_node(dn);
0135c482
CY		 956 if (err) {
957 f2fs_put_page(page, 1);
7735730d 958 return err;
0135c482 959 }
7735730d 960
e05df3b1
JK		 961 return 1;
962}
963
964static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs,
965 int ofs, int depth)
966{
e05df3b1
JK		 967 struct dnode_of_data rdn = *dn;
968 struct page *page;
969 struct f2fs_node *rn;
970 nid_t child_nid;
971 unsigned int child_nofs;
972 int freed = 0;
973 int i, ret;
974
975 if (dn->nid == 0)
976 return NIDS_PER_BLOCK + 1;
977
51dd6249
NJ		 978 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr);
979
4d57b86d 980 page = f2fs_get_node_page(F2FS_I_SB(dn->inode), dn->nid);
51dd6249
NJ		 981 if (IS_ERR(page)) {
982 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page));
e05df3b1 983 return PTR_ERR(page);
51dd6249 984 }
e05df3b1 985
4d57b86d 986 f2fs_ra_node_pages(page, ofs, NIDS_PER_BLOCK);
79344efb 987
45590710 988 rn = F2FS_NODE(page);
e05df3b1
JK		 989 if (depth < 3) {
990 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) {
991 child_nid = le32_to_cpu(rn->in.nid[i]);
992 if (child_nid == 0)
993 continue;
994 rdn.nid = child_nid;
995 ret = truncate_dnode(&rdn);
996 if (ret < 0)
997 goto out_err;
12719ae1
JK		 998 if (set_nid(page, i, 0, false))
999 dn->node_changed = true;
e05df3b1
JK		 1000 }
1001 } else {
1002 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1;
1003 for (i = ofs; i < NIDS_PER_BLOCK; i++) {
1004 child_nid = le32_to_cpu(rn->in.nid[i]);
1005 if (child_nid == 0) {
1006 child_nofs += NIDS_PER_BLOCK + 1;
1007 continue;
1008 }
1009 rdn.nid = child_nid;
1010 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1);
1011 if (ret == (NIDS_PER_BLOCK + 1)) {
12719ae1
JK		 1012 if (set_nid(page, i, 0, false))
1013 dn->node_changed = true;
e05df3b1
JK		 1014 child_nofs += ret;
1015 } else if (ret < 0 && ret != -ENOENT) {
1016 goto out_err;
1017 }
1018 }
1019 freed = child_nofs;
1020 }
1021
1022 if (!ofs) {
1023 /* remove current indirect node */
1024 dn->node_page = page;
7735730d
CY		 1025 ret = truncate_node(dn);
1026 if (ret)
1027 goto out_err;
e05df3b1
JK		 1028 freed++;
1029 } else {
1030 f2fs_put_page(page, 1);
1031 }
51dd6249 1032 trace_f2fs_truncate_nodes_exit(dn->inode, freed);
e05df3b1
JK		 1033 return freed;
1034
1035out_err:
1036 f2fs_put_page(page, 1);
51dd6249 1037 trace_f2fs_truncate_nodes_exit(dn->inode, ret);
e05df3b1
JK		 1038 return ret;
1039}
1040
1041static int truncate_partial_nodes(struct dnode_of_data *dn,
1042 struct f2fs_inode *ri, int *offset, int depth)
1043{
e05df3b1
JK		 1044 struct page *pages[2];
1045 nid_t nid[3];
1046 nid_t child_nid;
1047 int err = 0;
1048 int i;
1049 int idx = depth - 2;
1050
1051 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
1052 if (!nid[0])
1053 return 0;
1054
1055 /* get indirect nodes in the path */
a225dca3 1056 for (i = 0; i < idx + 1; i++) {
e1c42045 1057 /* reference count'll be increased */
4d57b86d 1058 pages[i] = f2fs_get_node_page(F2FS_I_SB(dn->inode), nid[i]);
e05df3b1 1059 if (IS_ERR(pages[i])) {
e05df3b1 1060 err = PTR_ERR(pages[i]);
a225dca3 1061 idx = i - 1;
e05df3b1
JK		 1062 goto fail;
1063 }
1064 nid[i + 1] = get_nid(pages[i], offset[i + 1], false);
1065 }
1066
4d57b86d 1067 f2fs_ra_node_pages(pages[idx], offset[idx + 1], NIDS_PER_BLOCK);
79344efb 1068
e05df3b1 1069 /* free direct nodes linked to a partial indirect node */
a225dca3 1070 for (i = offset[idx + 1]; i < NIDS_PER_BLOCK; i++) {
e05df3b1
JK		 1071 child_nid = get_nid(pages[idx], i, false);
1072 if (!child_nid)
1073 continue;
1074 dn->nid = child_nid;
1075 err = truncate_dnode(dn);
1076 if (err < 0)
1077 goto fail;
12719ae1
JK		 1078 if (set_nid(pages[idx], i, 0, false))
1079 dn->node_changed = true;
e05df3b1
JK		 1080 }
1081
a225dca3 1082 if (offset[idx + 1] == 0) {
e05df3b1
JK		 1083 dn->node_page = pages[idx];
1084 dn->nid = nid[idx];
7735730d
CY		 1085 err = truncate_node(dn);
1086 if (err)
1087 goto fail;
e05df3b1
JK		 1088 } else {
1089 f2fs_put_page(pages[idx], 1);
1090 }
1091 offset[idx]++;
a225dca3 1092 offset[idx + 1] = 0;
1093 idx--;
e05df3b1 1094fail:
a225dca3 1095 for (i = idx; i >= 0; i--)
e05df3b1 1096 f2fs_put_page(pages[i], 1);
51dd6249
NJ		 1097
1098 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err);
1099
e05df3b1
JK		 1100 return err;
1101}
1102
0a8165d7 1103/*
e05df3b1
JK		 1104 * All the block addresses of data and nodes should be nullified.
1105 */
4d57b86d 1106int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from)
e05df3b1 1107{
4081363f 1108 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
e05df3b1
JK		 1109 int err = 0, cont = 1;
1110 int level, offset[4], noffset[4];
7dd690c8 1111 unsigned int nofs = 0;
58bfaf44 1112 struct f2fs_inode *ri;
e05df3b1
JK		 1113 struct dnode_of_data dn;
1114 struct page *page;
1115
51dd6249
NJ		 1116 trace_f2fs_truncate_inode_blocks_enter(inode, from);
1117
81ca7350 1118 level = get_node_path(inode, from, offset, noffset);
9039d835
YF		 1119 if (level < 0) {
1120 trace_f2fs_truncate_inode_blocks_exit(inode, level);
adb6dc19 1121 return level;
9039d835 1122 }
ff373558 1123
4d57b86d 1124 page = f2fs_get_node_page(sbi, inode->i_ino);
51dd6249
NJ		 1125 if (IS_ERR(page)) {
1126 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page));
e05df3b1 1127 return PTR_ERR(page);
51dd6249 1128 }
e05df3b1
JK		 1129
1130 set_new_dnode(&dn, inode, page, NULL, 0);
1131 unlock_page(page);
1132
58bfaf44 1133 ri = F2FS_INODE(page);
e05df3b1
JK		 1134 switch (level) {
1135 case 0:
1136 case 1:
1137 nofs = noffset[1];
1138 break;
1139 case 2:
1140 nofs = noffset[1];
1141 if (!offset[level - 1])
1142 goto skip_partial;
58bfaf44 1143 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 1144 if (err < 0 && err != -ENOENT)
1145 goto fail;
1146 nofs += 1 + NIDS_PER_BLOCK;
1147 break;
1148 case 3:
1149 nofs = 5 + 2 * NIDS_PER_BLOCK;
1150 if (!offset[level - 1])
1151 goto skip_partial;
58bfaf44 1152 err = truncate_partial_nodes(&dn, ri, offset, level);
e05df3b1
JK		 1153 if (err < 0 && err != -ENOENT)
1154 goto fail;
1155 break;
1156 default:
1157 BUG();
1158 }
1159
1160skip_partial:
1161 while (cont) {
58bfaf44 1162 dn.nid = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]);
e05df3b1
JK		 1163 switch (offset[0]) {
1164 case NODE_DIR1_BLOCK:
1165 case NODE_DIR2_BLOCK:
1166 err = truncate_dnode(&dn);
1167 break;
1168
1169 case NODE_IND1_BLOCK:
1170 case NODE_IND2_BLOCK:
1171 err = truncate_nodes(&dn, nofs, offset[1], 2);
1172 break;
1173
1174 case NODE_DIND_BLOCK:
1175 err = truncate_nodes(&dn, nofs, offset[1], 3);
1176 cont = 0;
1177 break;
1178
1179 default:
1180 BUG();
1181 }
1182 if (err < 0 && err != -ENOENT)
1183 goto fail;
1184 if (offset[1] == 0 &&
58bfaf44 1185 ri->i_nid[offset[0] - NODE_DIR1_BLOCK]) {
e05df3b1 1186 lock_page(page);
ff373558 1187 BUG_ON(page->mapping != NODE_MAPPING(sbi));
bae0ee7a 1188 f2fs_wait_on_page_writeback(page, NODE, true, true);
58bfaf44 1189 ri->i_nid[offset[0] - NODE_DIR1_BLOCK] = 0;
e05df3b1
JK		 1190 set_page_dirty(page);
1191 unlock_page(page);
1192 }
1193 offset[1] = 0;
1194 offset[0]++;
1195 nofs += err;
1196 }
1197fail:
1198 f2fs_put_page(page, 0);
51dd6249 1199 trace_f2fs_truncate_inode_blocks_exit(inode, err);
e05df3b1
JK		 1200 return err > 0 ? 0 : err;
1201}
1202
9c77f754 1203/* caller must lock inode page */
4d57b86d 1204int f2fs_truncate_xattr_node(struct inode *inode)
4f16fb0f 1205{
4081363f 1206 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
4f16fb0f
JK		 1207 nid_t nid = F2FS_I(inode)->i_xattr_nid;
1208 struct dnode_of_data dn;
1209 struct page *npage;
7735730d 1210 int err;
4f16fb0f
JK		 1211
1212 if (!nid)
1213 return 0;
1214
4d57b86d 1215 npage = f2fs_get_node_page(sbi, nid);
4f16fb0f
JK		 1216 if (IS_ERR(npage))
1217 return PTR_ERR(npage);
1218
7735730d
CY		 1219 set_new_dnode(&dn, inode, NULL, npage, nid);
1220 err = truncate_node(&dn);
1221 if (err) {
1222 f2fs_put_page(npage, 1);
1223 return err;
1224 }
1225
205b9822 1226 f2fs_i_xnid_write(inode, 0);
65985d93 1227
4f16fb0f
JK		 1228 return 0;
1229}
1230
39936837 1231/*
4f4124d0
CY		 1232 * Caller should grab and release a rwsem by calling f2fs_lock_op() and
1233 * f2fs_unlock_op().
39936837 1234 */
4d57b86d 1235int f2fs_remove_inode_page(struct inode *inode)
e05df3b1 1236{
e05df3b1 1237 struct dnode_of_data dn;
13ec7297 1238 int err;
e05df3b1 1239
c2e69583 1240 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
4d57b86d 1241 err = f2fs_get_dnode_of_data(&dn, 0, LOOKUP_NODE);
13ec7297
CY		 1242 if (err)
1243 return err;
e05df3b1 1244
4d57b86d 1245 err = f2fs_truncate_xattr_node(inode);
13ec7297 1246 if (err) {
c2e69583 1247 f2fs_put_dnode(&dn);
13ec7297 1248 return err;
e05df3b1 1249 }
c2e69583
JK		 1250
1251 /* remove potential inline_data blocks */
1252 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
1253 S_ISLNK(inode->i_mode))
4d57b86d 1254 f2fs_truncate_data_blocks_range(&dn, 1);
c2e69583 1255
e1c42045 1256 /* 0 is possible, after f2fs_new_inode() has failed */
8d714f8a
JK		 1257 if (unlikely(f2fs_cp_error(F2FS_I_SB(inode)))) {
1258 f2fs_put_dnode(&dn);
1259 return -EIO;
1260 }
8b6810f8
CY		 1261
1262 if (unlikely(inode->i_blocks != 0 && inode->i_blocks != 8)) {
097a7686
CY		 1263 f2fs_warn(F2FS_I_SB(inode),
1264 "f2fs_remove_inode_page: inconsistent i_blocks, ino:%lu, iblocks:%llu",
1265 inode->i_ino, (unsigned long long)inode->i_blocks);
8b6810f8
CY		 1266 set_sbi_flag(F2FS_I_SB(inode), SBI_NEED_FSCK);
1267 }
c2e69583
JK		 1268
1269 /* will put inode & node pages */
7735730d
CY		 1270 err = truncate_node(&dn);
1271 if (err) {
1272 f2fs_put_dnode(&dn);
1273 return err;
1274 }
13ec7297 1275 return 0;
e05df3b1
JK		 1276}
1277
4d57b86d 1278struct page *f2fs_new_inode_page(struct inode *inode)
e05df3b1 1279{
e05df3b1
JK		 1280 struct dnode_of_data dn;
1281
1282 /* allocate inode page for new inode */
1283 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino);
44a83ff6
JK		 1284
1285 /* caller should f2fs_put_page(page, 1); */
4d57b86d 1286 return f2fs_new_node_page(&dn, 0);
e05df3b1
JK		 1287}
1288
4d57b86d 1289struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs)
e05df3b1 1290{
4081363f 1291 struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
25cc5d3b 1292 struct node_info new_ni;
e05df3b1
JK		 1293 struct page *page;
1294 int err;
1295
91942321 1296 if (unlikely(is_inode_flag_set(dn->inode, FI_NO_ALLOC)))
e05df3b1
JK		 1297 return ERR_PTR(-EPERM);
1298
300e129c 1299 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), dn->nid, false);
e05df3b1
JK		 1300 if (!page)
1301 return ERR_PTR(-ENOMEM);
1302
0abd675e 1303 if (unlikely((err = inc_valid_node_count(sbi, dn->inode, !ofs))))
9c02740c 1304 goto fail;
0abd675e 1305
25cc5d3b 1306#ifdef CONFIG_F2FS_CHECK_FS
a9419b63 1307 err = f2fs_get_node_info(sbi, dn->nid, &new_ni, false);
7735730d
CY		 1308 if (err) {
1309 dec_valid_node_count(sbi, dn->inode, !ofs);
1310 goto fail;
1311 }
141170b7
CY		 1312 if (unlikely(new_ni.blk_addr != NULL_ADDR)) {
1313 err = -EFSCORRUPTED;
1314 set_sbi_flag(sbi, SBI_NEED_FSCK);
95fa90c9 1315 f2fs_handle_error(sbi, ERROR_INVALID_BLKADDR);
141170b7
CY		 1316 goto fail;
1317 }
25cc5d3b
JK		 1318#endif
1319 new_ni.nid = dn->nid;
e05df3b1 1320 new_ni.ino = dn->inode->i_ino;
25cc5d3b
JK		 1321 new_ni.blk_addr = NULL_ADDR;
1322 new_ni.flag = 0;
1323 new_ni.version = 0;
479f40c4 1324 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
9c02740c 1325
bae0ee7a 1326 f2fs_wait_on_page_writeback(page, NODE, true, true);
9c02740c 1327 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true);
c5667575 1328 set_cold_node(page, S_ISDIR(dn->inode->i_mode));
237c0790
JK		 1329 if (!PageUptodate(page))
1330 SetPageUptodate(page);
12719ae1
JK		 1331 if (set_page_dirty(page))
1332 dn->node_changed = true;
e05df3b1 1333
4bc8e9bc 1334 if (f2fs_has_xattr_block(ofs))
205b9822 1335 f2fs_i_xnid_write(dn->inode, dn->nid);
479bd73a 1336
e05df3b1
JK		 1337 if (ofs == 0)
1338 inc_valid_inode_count(sbi);
e05df3b1
JK		 1339 return page;
1340
1341fail:
71e9fec5 1342 clear_node_page_dirty(page);
e05df3b1
JK		 1343 f2fs_put_page(page, 1);
1344 return ERR_PTR(err);
1345}
1346
56ae674c
JK		 1347/*
1348 * Caller should do after getting the following values.
1349 * 0: f2fs_put_page(page, 0)
86531d6b 1350 * LOCKED_PAGE or error: f2fs_put_page(page, 1)
56ae674c 1351 */
7649c873 1352static int read_node_page(struct page *page, blk_opf_t op_flags)
e05df3b1 1353{
4081363f 1354 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
e05df3b1 1355 struct node_info ni;
cf04e8eb 1356 struct f2fs_io_info fio = {
05ca3632 1357 .sbi = sbi,
cf04e8eb 1358 .type = NODE,
04d328de
MC		 1359 .op = REQ_OP_READ,
1360 .op_flags = op_flags,
05ca3632 1361 .page = page,
4375a336 1362 .encrypted_page = NULL,
cf04e8eb 1363 };
7735730d 1364 int err;
e05df3b1 1365
54c55c4e 1366 if (PageUptodate(page)) {
b42b179b
CY		 1367 if (!f2fs_inode_chksum_verify(sbi, page)) {
1368 ClearPageUptodate(page);
10f966bb 1369 return -EFSBADCRC;
b42b179b 1370 }
3bdad3c7 1371 return LOCKED_PAGE;
54c55c4e 1372 }
3bdad3c7 1373
a9419b63 1374 err = f2fs_get_node_info(sbi, page->index, &ni, false);
7735730d
CY		 1375 if (err)
1376 return err;
e05df3b1 1377
b7ec2061 1378 /* NEW_ADDR can be seen, after cp_error drops some dirty node pages */
e6ecb142 1379 if (unlikely(ni.blk_addr == NULL_ADDR || ni.blk_addr == NEW_ADDR)) {
2bca1e23 1380 ClearPageUptodate(page);
e05df3b1 1381 return -ENOENT;
393ff91f
JK		 1382 }
1383
7a9d7548 1384 fio.new_blkaddr = fio.old_blkaddr = ni.blk_addr;
8b83ac81
CY		 1385
1386 err = f2fs_submit_page_bio(&fio);
1387
1388 if (!err)
34a23525 1389 f2fs_update_iostat(sbi, NULL, FS_NODE_READ_IO, F2FS_BLKSIZE);
8b83ac81
CY		 1390
1391 return err;
e05df3b1
JK		 1392}
1393
0a8165d7 1394/*
e05df3b1
JK		 1395 * Readahead a node page
1396 */
4d57b86d 1397void f2fs_ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 1398{
e05df3b1 1399 struct page *apage;
56ae674c 1400 int err;
e05df3b1 1401
e8458725
CY		 1402 if (!nid)
1403 return;
4d57b86d 1404 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1405 return;
e8458725 1406
5ec2d99d 1407 apage = xa_load(&NODE_MAPPING(sbi)->i_pages, nid);
999270de 1408 if (apage)
393ff91f 1409 return;
e05df3b1 1410
300e129c 1411 apage = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1412 if (!apage)
1413 return;
1414
70246286 1415 err = read_node_page(apage, REQ_RAHEAD);
86531d6b 1416 f2fs_put_page(apage, err ? 1 : 0);
e05df3b1
JK		 1417}
1418
17a0ee55 1419static struct page *__get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid,
0e022ea8 1420 struct page *parent, int start)
e05df3b1 1421{
e05df3b1 1422 struct page *page;
0e022ea8 1423 int err;
e05df3b1 1424
e05df3b1
JK		 1425 if (!nid)
1426 return ERR_PTR(-ENOENT);
4d57b86d 1427 if (f2fs_check_nid_range(sbi, nid))
a4f843bd 1428 return ERR_PTR(-EINVAL);
afcb7ca0 1429repeat:
300e129c 1430 page = f2fs_grab_cache_page(NODE_MAPPING(sbi), nid, false);
e05df3b1
JK		 1431 if (!page)
1432 return ERR_PTR(-ENOMEM);
1433
70fd7614 1434 err = read_node_page(page, 0);
86531d6b 1435 if (err < 0) {
a7b8618a 1436 goto out_put_err;
86531d6b 1437 } else if (err == LOCKED_PAGE) {
1f258ec1 1438 err = 0;
56ae674c 1439 goto page_hit;
86531d6b 1440 }
e05df3b1 1441
0e022ea8 1442 if (parent)
4d57b86d 1443 f2fs_ra_node_pages(parent, start + 1, MAX_RA_NODE);
c718379b 1444
e05df3b1 1445 lock_page(page);
0e022ea8 1446
4ef51a8f 1447 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
afcb7ca0
JK		 1448 f2fs_put_page(page, 1);
1449 goto repeat;
1450 }
1563ac75 1451
1f258ec1
CY		 1452 if (unlikely(!PageUptodate(page))) {
1453 err = -EIO;
1563ac75 1454 goto out_err;
1f258ec1 1455 }
704956ec
CY		 1456
1457 if (!f2fs_inode_chksum_verify(sbi, page)) {
10f966bb 1458 err = -EFSBADCRC;
704956ec
CY		 1459 goto out_err;
1460 }
e0f56cb4 1461page_hit:
a7b8618a
JK		 1462 if (likely(nid == nid_of_node(page)))
1463 return page;
1464
1465 f2fs_warn(sbi, "inconsistent node block, nid:%lu, node_footer[nid:%u,ino:%u,ofs:%u,cpver:%llu,blkaddr:%u]",
dcbb4c10
JP		 1466 nid, nid_of_node(page), ino_of_node(page),
1467 ofs_of_node(page), cpver_of_node(page),
1468 next_blkaddr_of_node(page));
a7b8618a
JK		 1469 set_sbi_flag(sbi, SBI_NEED_FSCK);
1470 err = -EINVAL;
0c9df7fb 1471out_err:
a7b8618a
JK		 1472 ClearPageUptodate(page);
1473out_put_err:
82c7863e
JK		 1474 /* ENOENT comes from read_node_page which is not an error. */
1475 if (err != -ENOENT)
1476 f2fs_handle_page_eio(sbi, page->index, NODE);
a7b8618a
JK		 1477 f2fs_put_page(page, 1);
1478 return ERR_PTR(err);
e05df3b1
JK		 1479}
1480
4d57b86d 1481struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid)
0e022ea8
CY		 1482{
1483 return __get_node_page(sbi, nid, NULL, 0);
1484}
1485
4d57b86d 1486struct page *f2fs_get_node_page_ra(struct page *parent, int start)
0e022ea8
CY		 1487{
1488 struct f2fs_sb_info *sbi = F2FS_P_SB(parent);
1489 nid_t nid = get_nid(parent, start, false);
1490
1491 return __get_node_page(sbi, nid, parent, start);
1492}
1493
2049d4fc
JK		 1494static void flush_inline_data(struct f2fs_sb_info *sbi, nid_t ino)
1495{
1496 struct inode *inode;
1497 struct page *page;
0f3311a8 1498 int ret;
2049d4fc
JK		 1499
1500 /* should flush inline_data before evict_inode */
1501 inode = ilookup(sbi->sb, ino);
1502 if (!inode)
1503 return;
1504
01eccef7
CY		 1505 page = f2fs_pagecache_get_page(inode->i_mapping, 0,
1506 FGP_LOCK|FGP_NOWAIT, 0);
2049d4fc
JK		 1507 if (!page)
1508 goto iput_out;
1509
1510 if (!PageUptodate(page))
1511 goto page_out;
1512
1513 if (!PageDirty(page))
1514 goto page_out;
1515
1516 if (!clear_page_dirty_for_io(page))
1517 goto page_out;
1518
0f3311a8
CY		 1519 ret = f2fs_write_inline_data(inode, page);
1520 inode_dec_dirty_pages(inode);
4d57b86d 1521 f2fs_remove_dirty_inode(inode);
0f3311a8 1522 if (ret)
2049d4fc
JK		 1523 set_page_dirty(page);
1524page_out:
4a6de50d 1525 f2fs_put_page(page, 1);
2049d4fc
JK		 1526iput_out:
1527 iput(inode);
1528}
1529
608514de
JK		 1530static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
1531{
028a63a6 1532 pgoff_t index;
4f4a4f0f 1533 struct folio_batch fbatch;
608514de 1534 struct page *last_page = NULL;
4f4a4f0f 1535 int nr_folios;
608514de 1536
4f4a4f0f 1537 folio_batch_init(&fbatch);
608514de 1538 index = 0;
028a63a6 1539
4f4a4f0f
VMO		 1540 while ((nr_folios = filemap_get_folios_tag(NODE_MAPPING(sbi), &index,
1541 (pgoff_t)-1, PAGECACHE_TAG_DIRTY,
1542 &fbatch))) {
028a63a6 1543 int i;
608514de 1544
4f4a4f0f
VMO		 1545 for (i = 0; i < nr_folios; i++) {
1546 struct page *page = &fbatch.folios[i]->page;
608514de
JK		 1547
1548 if (unlikely(f2fs_cp_error(sbi))) {
1549 f2fs_put_page(last_page, 0);
4f4a4f0f 1550 folio_batch_release(&fbatch);
608514de
JK		 1551 return ERR_PTR(-EIO);
1552 }
1553
1554 if (!IS_DNODE(page) || !is_cold_node(page))
1555 continue;
1556 if (ino_of_node(page) != ino)
1557 continue;
1558
1559 lock_page(page);
1560
1561 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1562continue_unlock:
1563 unlock_page(page);
1564 continue;
1565 }
1566 if (ino_of_node(page) != ino)
1567 goto continue_unlock;
1568
1569 if (!PageDirty(page)) {
1570 /* someone wrote it for us */
1571 goto continue_unlock;
1572 }
1573
1574 if (last_page)
1575 f2fs_put_page(last_page, 0);
1576
1577 get_page(page);
1578 last_page = page;
1579 unlock_page(page);
1580 }
4f4a4f0f 1581 folio_batch_release(&fbatch);
608514de
JK		 1582 cond_resched();
1583 }
1584 return last_page;
1585}
1586
d68f735b 1587static int __write_node_page(struct page *page, bool atomic, bool *submitted,
b0af6d49 1588 struct writeback_control *wbc, bool do_balance,
50fa53ec 1589 enum iostat_type io_type, unsigned int *seq_id)
faa24895
JK		 1590{
1591 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1592 nid_t nid;
1593 struct node_info ni;
1594 struct f2fs_io_info fio = {
1595 .sbi = sbi,
39d787be 1596 .ino = ino_of_node(page),
faa24895
JK		 1597 .type = NODE,
1598 .op = REQ_OP_WRITE,
1599 .op_flags = wbc_to_write_flags(wbc),
1600 .page = page,
1601 .encrypted_page = NULL,
2eae077e 1602 .submitted = 0,
b0af6d49 1603 .io_type = io_type,
578c6478 1604 .io_wbc = wbc,
faa24895 1605 };
50fa53ec 1606 unsigned int seq;
faa24895
JK		 1607
1608 trace_f2fs_writepage(page, NODE);
1609
6d7c865c 1610 if (unlikely(f2fs_cp_error(sbi))) {
b62e71be
CY		 1611 /* keep node pages in remount-ro mode */
1612 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY)
1613 goto redirty_out;
28607bf3
JK		 1614 ClearPageUptodate(page);
1615 dec_page_count(sbi, F2FS_DIRTY_NODES);
1616 unlock_page(page);
1617 return 0;
6d7c865c 1618 }
db198ae0 1619
faa24895
JK		 1620 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1621 goto redirty_out;
faa24895 1622
100c0655
JK		 1623 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
1624 wbc->sync_mode == WB_SYNC_NONE &&
fd8c8caf
CY		 1625 IS_DNODE(page) && is_cold_node(page))
1626 goto redirty_out;
1627
faa24895
JK		 1628 /* get old block addr of this node page */
1629 nid = nid_of_node(page);
1630 f2fs_bug_on(sbi, page->index != nid);
1631
a9419b63 1632 if (f2fs_get_node_info(sbi, nid, &ni, !do_balance))
7735730d
CY		 1633 goto redirty_out;
1634
faa24895 1635 if (wbc->for_reclaim) {
e4544b63 1636 if (!f2fs_down_read_trylock(&sbi->node_write))
faa24895
JK		 1637 goto redirty_out;
1638 } else {
e4544b63 1639 f2fs_down_read(&sbi->node_write);
faa24895
JK		 1640 }
1641
faa24895
JK		 1642 /* This page is already truncated */
1643 if (unlikely(ni.blk_addr == NULL_ADDR)) {
1644 ClearPageUptodate(page);
1645 dec_page_count(sbi, F2FS_DIRTY_NODES);
e4544b63 1646 f2fs_up_read(&sbi->node_write);
faa24895
JK		 1647 unlock_page(page);
1648 return 0;
1649 }
1650
c9b60788 1651 if (__is_valid_data_blkaddr(ni.blk_addr) &&
93770ab7
CY		 1652 !f2fs_is_valid_blkaddr(sbi, ni.blk_addr,
1653 DATA_GENERIC_ENHANCE)) {
e4544b63 1654 f2fs_up_read(&sbi->node_write);
c9b60788 1655 goto redirty_out;
89d13c38 1656 }
c9b60788 1657
c550e25b 1658 if (atomic && !test_opt(sbi, NOBARRIER) && !f2fs_sb_has_blkzoned(sbi))
e7c75ab0
JK		 1659 fio.op_flags |= REQ_PREFLUSH | REQ_FUA;
1660
dc5a9412 1661 /* should add to global list before clearing PAGECACHE status */
50fa53ec
CY		 1662 if (f2fs_in_warm_node_list(sbi, page)) {
1663 seq = f2fs_add_fsync_node_entry(sbi, page);
1664 if (seq_id)
1665 *seq_id = seq;
1666 }
1667
dc5a9412 1668 set_page_writeback(page);
dc5a9412 1669
faa24895 1670 fio.old_blkaddr = ni.blk_addr;
4d57b86d 1671 f2fs_do_write_node_page(nid, &fio);
faa24895
JK		 1672 set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page));
1673 dec_page_count(sbi, F2FS_DIRTY_NODES);
e4544b63 1674 f2fs_up_read(&sbi->node_write);
faa24895 1675
d68f735b 1676 if (wbc->for_reclaim) {
bab475c5 1677 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, NODE);
d68f735b
JK		 1678 submitted = NULL;
1679 }
faa24895
JK		 1680
1681 unlock_page(page);
1682
d68f735b 1683 if (unlikely(f2fs_cp_error(sbi))) {
b9109b0e 1684 f2fs_submit_merged_write(sbi, NODE);
d68f735b
JK		 1685 submitted = NULL;
1686 }
1687 if (submitted)
1688 *submitted = fio.submitted;
faa24895 1689
401db79f
YS		 1690 if (do_balance)
1691 f2fs_balance_fs(sbi, false);
faa24895
JK		 1692 return 0;
1693
1694redirty_out:
1695 redirty_page_for_writepage(wbc, page);
1696 return AOP_WRITEPAGE_ACTIVATE;
1697}
1698
48018b4c 1699int f2fs_move_node_page(struct page *node_page, int gc_type)
f15194fc 1700{
48018b4c
CY		 1701 int err = 0;
1702
f15194fc
YH		 1703 if (gc_type == FG_GC) {
1704 struct writeback_control wbc = {
1705 .sync_mode = WB_SYNC_ALL,
1706 .nr_to_write = 1,
1707 .for_reclaim = 0,
1708 };
1709
bae0ee7a 1710 f2fs_wait_on_page_writeback(node_page, NODE, true, true);
8d64d365
CY		 1711
1712 set_page_dirty(node_page);
1713
48018b4c
CY		 1714 if (!clear_page_dirty_for_io(node_page)) {
1715 err = -EAGAIN;
f15194fc 1716 goto out_page;
48018b4c 1717 }
f15194fc
YH		 1718
1719 if (__write_node_page(node_page, false, NULL,
48018b4c
CY		 1720 &wbc, false, FS_GC_NODE_IO, NULL)) {
1721 err = -EAGAIN;
f15194fc 1722 unlock_page(node_page);
48018b4c 1723 }
f15194fc
YH		 1724 goto release_page;
1725 } else {
1726 /* set page dirty and write it */
1727 if (!PageWriteback(node_page))
1728 set_page_dirty(node_page);
1729 }
1730out_page:
1731 unlock_page(node_page);
1732release_page:
1733 f2fs_put_page(node_page, 0);
48018b4c 1734 return err;
f15194fc
YH		 1735}
1736
faa24895
JK		 1737static int f2fs_write_node_page(struct page *page,
1738 struct writeback_control *wbc)
1739{
50fa53ec
CY		 1740 return __write_node_page(page, false, NULL, wbc, false,
1741 FS_NODE_IO, NULL);
faa24895
JK		 1742}
1743
4d57b86d 1744int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode,
50fa53ec
CY		 1745 struct writeback_control *wbc, bool atomic,
1746 unsigned int *seq_id)
e05df3b1 1747{
028a63a6 1748 pgoff_t index;
e6e46e1e 1749 struct folio_batch fbatch;
c267ec15 1750 int ret = 0;
608514de
JK		 1751 struct page *last_page = NULL;
1752 bool marked = false;
26de9b11 1753 nid_t ino = inode->i_ino;
e6e46e1e 1754 int nr_folios;
bab475c5 1755 int nwritten = 0;
52681375 1756
608514de
JK		 1757 if (atomic) {
1758 last_page = last_fsync_dnode(sbi, ino);
1759 if (IS_ERR_OR_NULL(last_page))
1760 return PTR_ERR_OR_ZERO(last_page);
1761 }
1762retry:
e6e46e1e 1763 folio_batch_init(&fbatch);
52681375 1764 index = 0;
028a63a6 1765
e6e46e1e
VMO		 1766 while ((nr_folios = filemap_get_folios_tag(NODE_MAPPING(sbi), &index,
1767 (pgoff_t)-1, PAGECACHE_TAG_DIRTY,
1768 &fbatch))) {
028a63a6 1769 int i;
52681375 1770
e6e46e1e
VMO		 1771 for (i = 0; i < nr_folios; i++) {
1772 struct page *page = &fbatch.folios[i]->page;
d68f735b 1773 bool submitted = false;
52681375
JK		 1774
1775 if (unlikely(f2fs_cp_error(sbi))) {
608514de 1776 f2fs_put_page(last_page, 0);
e6e46e1e 1777 folio_batch_release(&fbatch);
9de69279
CY		 1778 ret = -EIO;
1779 goto out;
52681375
JK		 1780 }
1781
1782 if (!IS_DNODE(page) || !is_cold_node(page))
1783 continue;
1784 if (ino_of_node(page) != ino)
1785 continue;
1786
1787 lock_page(page);
1788
1789 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1790continue_unlock:
1791 unlock_page(page);
1792 continue;
1793 }
1794 if (ino_of_node(page) != ino)
1795 goto continue_unlock;
1796
608514de 1797 if (!PageDirty(page) && page != last_page) {
52681375
JK		 1798 /* someone wrote it for us */
1799 goto continue_unlock;
1800 }
1801
bae0ee7a 1802 f2fs_wait_on_page_writeback(page, NODE, true, true);
52681375 1803
d29fd172
JK		 1804 set_fsync_mark(page, 0);
1805 set_dentry_mark(page, 0);
1806
608514de
JK		 1807 if (!atomic || page == last_page) {
1808 set_fsync_mark(page, 1);
47c8ebcc 1809 percpu_counter_inc(&sbi->rf_node_block_count);
26de9b11
JK		 1810 if (IS_INODE(page)) {
1811 if (is_inode_flag_set(inode,
1812 FI_DIRTY_INODE))
4d57b86d 1813 f2fs_update_inode(inode, page);
608514de 1814 set_dentry_mark(page,
4d57b86d 1815 f2fs_need_dentry_mark(sbi, ino));
26de9b11 1816 }
a87aff1d 1817 /* may be written by other thread */
608514de
JK		 1818 if (!PageDirty(page))
1819 set_page_dirty(page);
1820 }
1821
1822 if (!clear_page_dirty_for_io(page))
1823 goto continue_unlock;
52681375 1824
e7c75ab0 1825 ret = __write_node_page(page, atomic &&
d68f735b 1826 page == last_page,
b0af6d49 1827 &submitted, wbc, true,
50fa53ec 1828 FS_NODE_IO, seq_id);
c267ec15 1829 if (ret) {
52681375 1830 unlock_page(page);
608514de
JK		 1831 f2fs_put_page(last_page, 0);
1832 break;
d68f735b 1833 } else if (submitted) {
bab475c5 1834 nwritten++;
608514de 1835 }
3f5f4959 1836
608514de
JK		 1837 if (page == last_page) {
1838 f2fs_put_page(page, 0);
1839 marked = true;
52681375 1840 break;
c267ec15 1841 }
52681375 1842 }
e6e46e1e 1843 folio_batch_release(&fbatch);
52681375
JK		 1844 cond_resched();
1845
608514de 1846 if (ret || marked)
52681375
JK		 1847 break;
1848 }
608514de 1849 if (!ret && atomic && !marked) {
dcbb4c10
JP		 1850 f2fs_debug(sbi, "Retry to write fsync mark: ino=%u, idx=%lx",
1851 ino, last_page->index);
608514de 1852 lock_page(last_page);
bae0ee7a 1853 f2fs_wait_on_page_writeback(last_page, NODE, true, true);
608514de
JK		 1854 set_page_dirty(last_page);
1855 unlock_page(last_page);
1856 goto retry;
1857 }
9de69279 1858out:
bab475c5
CY		 1859 if (nwritten)
1860 f2fs_submit_merged_write_cond(sbi, NULL, NULL, ino, NODE);
5f029c04 1861 return ret ? -EIO : 0;
52681375
JK		 1862}
1863
052a82d8
CY		 1864static int f2fs_match_ino(struct inode *inode, unsigned long ino, void *data)
1865{
1866 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1867 bool clean;
1868
1869 if (inode->i_ino != ino)
1870 return 0;
1871
1872 if (!is_inode_flag_set(inode, FI_DIRTY_INODE))
1873 return 0;
1874
1875 spin_lock(&sbi->inode_lock[DIRTY_META]);
1876 clean = list_empty(&F2FS_I(inode)->gdirty_list);
1877 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1878
1879 if (clean)
1880 return 0;
1881
1882 inode = igrab(inode);
1883 if (!inode)
1884 return 0;
1885 return 1;
1886}
1887
1888static bool flush_dirty_inode(struct page *page)
1889{
1890 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
1891 struct inode *inode;
1892 nid_t ino = ino_of_node(page);
1893
1894 inode = find_inode_nowait(sbi->sb, ino, f2fs_match_ino, NULL);
1895 if (!inode)
1896 return false;
1897
1898 f2fs_update_inode(inode, page);
1899 unlock_page(page);
1900
1901 iput(inode);
1902 return true;
1903}
1904
68e79baf 1905void f2fs_flush_inline_data(struct f2fs_sb_info *sbi)
34c061ad
SL		 1906{
1907 pgoff_t index = 0;
a40a4ad1
VMO		 1908 struct folio_batch fbatch;
1909 int nr_folios;
34c061ad 1910
a40a4ad1 1911 folio_batch_init(&fbatch);
34c061ad 1912
a40a4ad1
VMO		 1913 while ((nr_folios = filemap_get_folios_tag(NODE_MAPPING(sbi), &index,
1914 (pgoff_t)-1, PAGECACHE_TAG_DIRTY,
1915 &fbatch))) {
34c061ad
SL		 1916 int i;
1917
a40a4ad1
VMO		 1918 for (i = 0; i < nr_folios; i++) {
1919 struct page *page = &fbatch.folios[i]->page;
34c061ad
SL		 1920
1921 if (!IS_DNODE(page))
1922 continue;
1923
1924 lock_page(page);
1925
1926 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
1927continue_unlock:
1928 unlock_page(page);
1929 continue;
1930 }
1931
1932 if (!PageDirty(page)) {
1933 /* someone wrote it for us */
1934 goto continue_unlock;
1935 }
1936
1937 /* flush inline_data, if it's async context. */
b763f3be
CY		 1938 if (page_private_inline(page)) {
1939 clear_page_private_inline(page);
34c061ad
SL		 1940 unlock_page(page);
1941 flush_inline_data(sbi, ino_of_node(page));
1942 continue;
1943 }
1944 unlock_page(page);
1945 }
a40a4ad1 1946 folio_batch_release(&fbatch);
34c061ad
SL		 1947 cond_resched();
1948 }
34c061ad
SL		 1949}
1950
4d57b86d
CY		 1951int f2fs_sync_node_pages(struct f2fs_sb_info *sbi,
1952 struct writeback_control *wbc,
b0af6d49 1953 bool do_balance, enum iostat_type io_type)
52681375 1954{
028a63a6 1955 pgoff_t index;
7525486a 1956 struct folio_batch fbatch;
52681375 1957 int step = 0;
12bb0a8f 1958 int nwritten = 0;
3f5f4959 1959 int ret = 0;
7525486a 1960 int nr_folios, done = 0;
e05df3b1 1961
7525486a 1962 folio_batch_init(&fbatch);
e05df3b1
JK		 1963
1964next_step:
1965 index = 0;
028a63a6 1966
7525486a
VMO		 1967 while (!done && (nr_folios = filemap_get_folios_tag(NODE_MAPPING(sbi),
1968 &index, (pgoff_t)-1, PAGECACHE_TAG_DIRTY,
1969 &fbatch))) {
028a63a6 1970 int i;
e05df3b1 1971
7525486a
VMO		 1972 for (i = 0; i < nr_folios; i++) {
1973 struct page *page = &fbatch.folios[i]->page;
d68f735b 1974 bool submitted = false;
e05df3b1 1975
c29fd0c0
CY		 1976 /* give a priority to WB_SYNC threads */
1977 if (atomic_read(&sbi->wb_sync_req[NODE]) &&
1978 wbc->sync_mode == WB_SYNC_NONE) {
1979 done = 1;
1980 break;
1981 }
1982
e05df3b1
JK		 1983 /*
1984 * flushing sequence with step:
1985 * 0. indirect nodes
1986 * 1. dentry dnodes
1987 * 2. file dnodes
1988 */
1989 if (step == 0 && IS_DNODE(page))
1990 continue;
1991 if (step == 1 && (!IS_DNODE(page) ||
1992 is_cold_node(page)))
1993 continue;
1994 if (step == 2 && (!IS_DNODE(page) ||
1995 !is_cold_node(page)))
1996 continue;
9a4cbc9e 1997lock_node:
4b270a8c
CY		 1998 if (wbc->sync_mode == WB_SYNC_ALL)
1999 lock_page(page);
2000 else if (!trylock_page(page))
e05df3b1
JK		 2001 continue;
2002
4ef51a8f 2003 if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
e05df3b1
JK		 2004continue_unlock:
2005 unlock_page(page);
2006 continue;
2007 }
e05df3b1
JK		 2008
2009 if (!PageDirty(page)) {
2010 /* someone wrote it for us */
2011 goto continue_unlock;
2012 }
2013
b0f3b87f
JK		 2014 /* flush inline_data/inode, if it's async context. */
2015 if (!do_balance)
2016 goto write_node;
2017
2018 /* flush inline_data */
b763f3be
CY		 2019 if (page_private_inline(page)) {
2020 clear_page_private_inline(page);
2049d4fc
JK		 2021 unlock_page(page);
2022 flush_inline_data(sbi, ino_of_node(page));
9a4cbc9e 2023 goto lock_node;
052a82d8
CY		 2024 }
2025
2026 /* flush dirty inode */
7859e97f
JK		 2027 if (IS_INODE(page) && flush_dirty_inode(page))
2028 goto lock_node;
b0f3b87f 2029write_node:
bae0ee7a 2030 f2fs_wait_on_page_writeback(page, NODE, true, true);
fa3d2bdf 2031
e05df3b1
JK		 2032 if (!clear_page_dirty_for_io(page))
2033 goto continue_unlock;
2034
52681375
JK		 2035 set_fsync_mark(page, 0);
2036 set_dentry_mark(page, 0);
52746519 2037
401db79f 2038 ret = __write_node_page(page, false, &submitted,
50fa53ec 2039 wbc, do_balance, io_type, NULL);
d68f735b 2040 if (ret)
52746519 2041 unlock_page(page);
d68f735b 2042 else if (submitted)
3f5f4959 2043 nwritten++;
e05df3b1
JK		 2044
2045 if (--wbc->nr_to_write == 0)
2046 break;
2047 }
7525486a 2048 folio_batch_release(&fbatch);
e05df3b1
JK		 2049 cond_resched();
2050
2051 if (wbc->nr_to_write == 0) {
2052 step = 2;
2053 break;
2054 }
2055 }
2056
2057 if (step < 2) {
100c0655
JK		 2058 if (!is_sbi_flag_set(sbi, SBI_CP_DISABLED) &&
2059 wbc->sync_mode == WB_SYNC_NONE && step == 1)
fd8c8caf 2060 goto out;
e05df3b1
JK		 2061 step++;
2062 goto next_step;
2063 }
fd8c8caf 2064out:
3f5f4959 2065 if (nwritten)
b9109b0e 2066 f2fs_submit_merged_write(sbi, NODE);
db198ae0
CY		 2067
2068 if (unlikely(f2fs_cp_error(sbi)))
2069 return -EIO;
3f5f4959 2070 return ret;
e05df3b1
JK		 2071}
2072
50fa53ec
CY		 2073int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi,
2074 unsigned int seq_id)
cfe58f9d 2075{
50fa53ec
CY		 2076 struct fsync_node_entry *fn;
2077 struct page *page;
2078 struct list_head *head = &sbi->fsync_node_list;
2079 unsigned long flags;
2080 unsigned int cur_seq_id = 0;
cfe58f9d 2081
50fa53ec
CY		 2082 while (seq_id && cur_seq_id < seq_id) {
2083 spin_lock_irqsave(&sbi->fsync_node_lock, flags);
2084 if (list_empty(head)) {
2085 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
2086 break;
2087 }
2088 fn = list_first_entry(head, struct fsync_node_entry, list);
2089 if (fn->seq_id > seq_id) {
2090 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
2091 break;
2092 }
2093 cur_seq_id = fn->seq_id;
2094 page = fn->page;
2095 get_page(page);
2096 spin_unlock_irqrestore(&sbi->fsync_node_lock, flags);
4ef51a8f 2097
bae0ee7a 2098 f2fs_wait_on_page_writeback(page, NODE, true, false);
cfe58f9d 2099
50fa53ec 2100 put_page(page);
cfe58f9d
JK		 2101 }
2102
08c3eab5 2103 return filemap_check_errors(NODE_MAPPING(sbi));
cfe58f9d
JK		 2104}
2105
e05df3b1
JK		 2106static int f2fs_write_node_pages(struct address_space *mapping,
2107 struct writeback_control *wbc)
2108{
4081363f 2109 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
9dfa1baf 2110 struct blk_plug plug;
50c8cdb3 2111 long diff;
e05df3b1 2112
0771fcc7
CY		 2113 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
2114 goto skip_write;
2115
4660f9c0 2116 /* balancing f2fs's metadata in background */
7bcd0cfa 2117 f2fs_balance_fs_bg(sbi, true);
e05df3b1 2118
a7fdffbd 2119 /* collect a number of dirty node pages and write together */
812a9597
JK		 2120 if (wbc->sync_mode != WB_SYNC_ALL &&
2121 get_pages(sbi, F2FS_DIRTY_NODES) <
2122 nr_pages_to_skip(sbi, NODE))
d3baf95d 2123 goto skip_write;
a7fdffbd 2124
c29fd0c0
CY		 2125 if (wbc->sync_mode == WB_SYNC_ALL)
2126 atomic_inc(&sbi->wb_sync_req[NODE]);
34415099
CY		 2127 else if (atomic_read(&sbi->wb_sync_req[NODE])) {
2128 /* to avoid potential deadlock */
2129 if (current->plug)
2130 blk_finish_plug(current->plug);
c29fd0c0 2131 goto skip_write;
34415099 2132 }
c29fd0c0 2133
d31c7c3f
YH		 2134 trace_f2fs_writepages(mapping->host, wbc, NODE);
2135
50c8cdb3 2136 diff = nr_pages_to_write(sbi, NODE, wbc);
9dfa1baf 2137 blk_start_plug(&plug);
4d57b86d 2138 f2fs_sync_node_pages(sbi, wbc, true, FS_NODE_IO);
9dfa1baf 2139 blk_finish_plug(&plug);
50c8cdb3 2140 wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
c29fd0c0
CY		 2141
2142 if (wbc->sync_mode == WB_SYNC_ALL)
2143 atomic_dec(&sbi->wb_sync_req[NODE]);
e05df3b1 2144 return 0;
d3baf95d
JK		 2145
2146skip_write:
2147 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_NODES);
d31c7c3f 2148 trace_f2fs_writepages(mapping->host, wbc, NODE);
d3baf95d 2149 return 0;
e05df3b1
JK		 2150}
2151
cbc975b1
MWO		 2152static bool f2fs_dirty_node_folio(struct address_space *mapping,
2153 struct folio *folio)
e05df3b1 2154{
cbc975b1 2155 trace_f2fs_set_page_dirty(&folio->page, NODE);
26c6b887 2156
cbc975b1
MWO		 2157 if (!folio_test_uptodate(folio))
2158 folio_mark_uptodate(folio);
54c55c4e 2159#ifdef CONFIG_F2FS_CHECK_FS
cbc975b1 2160 if (IS_INODE(&folio->page))
29c87793 2161 f2fs_inode_chksum_set(F2FS_M_SB(mapping), &folio->page);
54c55c4e 2162#endif
9b7eadd9 2163 if (filemap_dirty_folio(mapping, folio)) {
29c87793 2164 inc_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_NODES);
cbc975b1
MWO		 2165 set_page_private_reference(&folio->page);
2166 return true;
e05df3b1 2167 }
cbc975b1 2168 return false;
e05df3b1
JK		 2169}
2170
0a8165d7 2171/*
e05df3b1
JK		 2172 * Structure of the f2fs node operations
2173 */
2174const struct address_space_operations f2fs_node_aops = {
2175 .writepage = f2fs_write_node_page,
2176 .writepages = f2fs_write_node_pages,
cbc975b1 2177 .dirty_folio = f2fs_dirty_node_folio,
91503996 2178 .invalidate_folio = f2fs_invalidate_folio,
c26cd045 2179 .release_folio = f2fs_release_folio,
1d5b9bd6 2180 .migrate_folio = filemap_migrate_folio,
e05df3b1
JK		 2181};
2182
8a7ed66a
JK		 2183static struct free_nid *__lookup_free_nid_list(struct f2fs_nm_info *nm_i,
2184 nid_t n)
e05df3b1 2185{
8a7ed66a 2186 return radix_tree_lookup(&nm_i->free_nid_root, n);
e05df3b1
JK		 2187}
2188
9a4ffdf5 2189static int __insert_free_nid(struct f2fs_sb_info *sbi,
b815bdc7 2190 struct free_nid *i)
e05df3b1 2191{
b8559dc2 2192 struct f2fs_nm_info *nm_i = NM_I(sbi);
a0761f63 2193 int err = radix_tree_insert(&nm_i->free_nid_root, i->nid, i);
5f029c04 2194
a0761f63
FL		 2195 if (err)
2196 return err;
eb0aa4b8 2197
b815bdc7
LS		 2198 nm_i->nid_cnt[FREE_NID]++;
2199 list_add_tail(&i->list, &nm_i->free_nid_list);
eb0aa4b8 2200 return 0;
b8559dc2
CY		 2201}
2202
9a4ffdf5 2203static void __remove_free_nid(struct f2fs_sb_info *sbi,
a0761f63 2204 struct free_nid *i, enum nid_state state)
b8559dc2
CY		 2205{
2206 struct f2fs_nm_info *nm_i = NM_I(sbi);
2207
9a4ffdf5
CY		 2208 f2fs_bug_on(sbi, state != i->state);
2209 nm_i->nid_cnt[state]--;
2210 if (state == FREE_NID)
2211 list_del(&i->list);
a0761f63
FL		 2212 radix_tree_delete(&nm_i->free_nid_root, i->nid);
2213}
2214
2215static void __move_free_nid(struct f2fs_sb_info *sbi, struct free_nid *i,
2216 enum nid_state org_state, enum nid_state dst_state)
b8559dc2
CY		 2217{
2218 struct f2fs_nm_info *nm_i = NM_I(sbi);
2219
a0761f63
FL		 2220 f2fs_bug_on(sbi, org_state != i->state);
2221 i->state = dst_state;
2222 nm_i->nid_cnt[org_state]--;
2223 nm_i->nid_cnt[dst_state]++;
2224
2225 switch (dst_state) {
2226 case PREALLOC_NID:
2227 list_del(&i->list);
2228 break;
2229 case FREE_NID:
2230 list_add_tail(&i->list, &nm_i->free_nid_list);
2231 break;
2232 default:
2233 BUG_ON(1);
2234 }
e05df3b1
JK		 2235}
2236
94c821fb
CY		 2237bool f2fs_nat_bitmap_enabled(struct f2fs_sb_info *sbi)
2238{
2239 struct f2fs_nm_info *nm_i = NM_I(sbi);
2240 unsigned int i;
2241 bool ret = true;
2242
e4544b63 2243 f2fs_down_read(&nm_i->nat_tree_lock);
94c821fb
CY		 2244 for (i = 0; i < nm_i->nat_blocks; i++) {
2245 if (!test_bit_le(i, nm_i->nat_block_bitmap)) {
2246 ret = false;
2247 break;
2248 }
2249 }
e4544b63 2250 f2fs_up_read(&nm_i->nat_tree_lock);
94c821fb
CY		 2251
2252 return ret;
2253}
2254
5921aaa1
L		 2255static void update_free_nid_bitmap(struct f2fs_sb_info *sbi, nid_t nid,
2256 bool set, bool build)
2257{
2258 struct f2fs_nm_info *nm_i = NM_I(sbi);
2259 unsigned int nat_ofs = NAT_BLOCK_OFFSET(nid);
2260 unsigned int nid_ofs = nid - START_NID(nid);
2261
2262 if (!test_bit_le(nat_ofs, nm_i->nat_block_bitmap))
2263 return;
2264
2265 if (set) {
2266 if (test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2267 return;
2268 __set_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2269 nm_i->free_nid_count[nat_ofs]++;
2270 } else {
2271 if (!test_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]))
2272 return;
2273 __clear_bit_le(nid_ofs, nm_i->free_nid_bitmap[nat_ofs]);
2274 if (!build)
2275 nm_i->free_nid_count[nat_ofs]--;
2276 }
2277}
2278
4ac91242 2279/* return if the nid is recognized as free */
5921aaa1
L		 2280static bool add_free_nid(struct f2fs_sb_info *sbi,
2281 nid_t nid, bool build, bool update)
e05df3b1 2282{
6fb03f3a 2283 struct f2fs_nm_info *nm_i = NM_I(sbi);
30a61ddf 2284 struct free_nid *i, *e;
59bbd474 2285 struct nat_entry *ne;
30a61ddf
CY		 2286 int err = -EINVAL;
2287 bool ret = false;
9198aceb
JK		 2288
2289 /* 0 nid should not be used */
cfb271d4 2290 if (unlikely(nid == 0))
4ac91242 2291 return false;
59bbd474 2292
626bcf2b
CY		 2293 if (unlikely(f2fs_check_nid_range(sbi, nid)))
2294 return false;
2295
32410577 2296 i = f2fs_kmem_cache_alloc(free_nid_slab, GFP_NOFS, true, NULL);
e05df3b1 2297 i->nid = nid;
9a4ffdf5 2298 i->state = FREE_NID;
e05df3b1 2299
5921aaa1 2300 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
769ec6e5 2301
b8559dc2 2302 spin_lock(&nm_i->nid_list_lock);
30a61ddf
CY		 2303
2304 if (build) {
2305 /*
2306 * Thread A Thread B
2307 * - f2fs_create
2308 * - f2fs_new_inode
4d57b86d 2309 * - f2fs_alloc_nid
9a4ffdf5 2310 * - __insert_nid_to_list(PREALLOC_NID)
30a61ddf 2311 * - f2fs_balance_fs_bg
4d57b86d
CY		 2312 * - f2fs_build_free_nids
2313 * - __f2fs_build_free_nids
30a61ddf
CY		 2314 * - scan_nat_page
2315 * - add_free_nid
2316 * - __lookup_nat_cache
2317 * - f2fs_add_link
4d57b86d
CY		 2318 * - f2fs_init_inode_metadata
2319 * - f2fs_new_inode_page
2320 * - f2fs_new_node_page
30a61ddf 2321 * - set_node_addr
4d57b86d 2322 * - f2fs_alloc_nid_done
9a4ffdf5
CY		 2323 * - __remove_nid_from_list(PREALLOC_NID)
2324 * - __insert_nid_to_list(FREE_NID)
30a61ddf
CY		 2325 */
2326 ne = __lookup_nat_cache(nm_i, nid);
2327 if (ne && (!get_nat_flag(ne, IS_CHECKPOINTED) ||
2328 nat_get_blkaddr(ne) != NULL_ADDR))
2329 goto err_out;
2330
2331 e = __lookup_free_nid_list(nm_i, nid);
2332 if (e) {
9a4ffdf5 2333 if (e->state == FREE_NID)
30a61ddf
CY		 2334 ret = true;
2335 goto err_out;
2336 }
2337 }
2338 ret = true;
b815bdc7 2339 err = __insert_free_nid(sbi, i);
30a61ddf 2340err_out:
5921aaa1
L		 2341 if (update) {
2342 update_free_nid_bitmap(sbi, nid, ret, build);
2343 if (!build)
2344 nm_i->available_nids++;
2345 }
eb0aa4b8
JK		 2346 spin_unlock(&nm_i->nid_list_lock);
2347 radix_tree_preload_end();
5921aaa1 2348
30a61ddf 2349 if (err)
e05df3b1 2350 kmem_cache_free(free_nid_slab, i);
30a61ddf 2351 return ret;
e05df3b1
JK		 2352}
2353
b8559dc2 2354static void remove_free_nid(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2355{
b8559dc2 2356 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1 2357 struct free_nid *i;
cf0ee0f0
CY		 2358 bool need_free = false;
2359
b8559dc2 2360 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2361 i = __lookup_free_nid_list(nm_i, nid);
9a4ffdf5 2362 if (i && i->state == FREE_NID) {
a0761f63 2363 __remove_free_nid(sbi, i, FREE_NID);
cf0ee0f0 2364 need_free = true;
e05df3b1 2365 }
b8559dc2 2366 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2367
2368 if (need_free)
2369 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2370}
2371
e2374015 2372static int scan_nat_page(struct f2fs_sb_info *sbi,
e05df3b1
JK		 2373 struct page *nat_page, nid_t start_nid)
2374{
6fb03f3a 2375 struct f2fs_nm_info *nm_i = NM_I(sbi);
e05df3b1
JK		 2376 struct f2fs_nat_block *nat_blk = page_address(nat_page);
2377 block_t blk_addr;
4ac91242 2378 unsigned int nat_ofs = NAT_BLOCK_OFFSET(start_nid);
e05df3b1
JK		 2379 int i;
2380
23380b85 2381 __set_bit_le(nat_ofs, nm_i->nat_block_bitmap);
4ac91242 2382
e05df3b1
JK		 2383 i = start_nid % NAT_ENTRY_PER_BLOCK;
2384
2385 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) {
cfb271d4 2386 if (unlikely(start_nid >= nm_i->max_nid))
04431c44 2387 break;
23d38844
HL		 2388
2389 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr);
e2374015
CY		 2390
2391 if (blk_addr == NEW_ADDR)
2392 return -EINVAL;
2393
5921aaa1
L		 2394 if (blk_addr == NULL_ADDR) {
2395 add_free_nid(sbi, start_nid, true, true);
2396 } else {
2397 spin_lock(&NM_I(sbi)->nid_list_lock);
2398 update_free_nid_bitmap(sbi, start_nid, false, true);
2399 spin_unlock(&NM_I(sbi)->nid_list_lock);
2400 }
4ac91242 2401 }
e2374015
CY		 2402
2403 return 0;
4ac91242
CY		 2404}
2405
2fbaa25f 2406static void scan_curseg_cache(struct f2fs_sb_info *sbi)
4ac91242 2407{
4ac91242
CY		 2408 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2409 struct f2fs_journal *journal = curseg->journal;
2fbaa25f
CY		 2410 int i;
2411
2412 down_read(&curseg->journal_rwsem);
2413 for (i = 0; i < nats_in_cursum(journal); i++) {
2414 block_t addr;
2415 nid_t nid;
2416
2417 addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
2418 nid = le32_to_cpu(nid_in_journal(journal, i));
2419 if (addr == NULL_ADDR)
5921aaa1 2420 add_free_nid(sbi, nid, true, false);
2fbaa25f
CY		 2421 else
2422 remove_free_nid(sbi, nid);
2423 }
2424 up_read(&curseg->journal_rwsem);
2425}
2426
2427static void scan_free_nid_bits(struct f2fs_sb_info *sbi)
2428{
2429 struct f2fs_nm_info *nm_i = NM_I(sbi);
4ac91242 2430 unsigned int i, idx;
97456574 2431 nid_t nid;
4ac91242 2432
e4544b63 2433 f2fs_down_read(&nm_i->nat_tree_lock);
4ac91242
CY		 2434
2435 for (i = 0; i < nm_i->nat_blocks; i++) {
2436 if (!test_bit_le(i, nm_i->nat_block_bitmap))
2437 continue;
586d1492
CY		 2438 if (!nm_i->free_nid_count[i])
2439 continue;
4ac91242 2440 for (idx = 0; idx < NAT_ENTRY_PER_BLOCK; idx++) {
97456574
FL		 2441 idx = find_next_bit_le(nm_i->free_nid_bitmap[i],
2442 NAT_ENTRY_PER_BLOCK, idx);
2443 if (idx >= NAT_ENTRY_PER_BLOCK)
2444 break;
4ac91242
CY		 2445
2446 nid = i * NAT_ENTRY_PER_BLOCK + idx;
5921aaa1 2447 add_free_nid(sbi, nid, true, false);
4ac91242 2448
9a4ffdf5 2449 if (nm_i->nid_cnt[FREE_NID] >= MAX_FREE_NIDS)
4ac91242
CY		 2450 goto out;
2451 }
2452 }
2453out:
2fbaa25f 2454 scan_curseg_cache(sbi);
4ac91242 2455
e4544b63 2456 f2fs_up_read(&nm_i->nat_tree_lock);
e05df3b1
JK		 2457}
2458
e2374015 2459static int __f2fs_build_free_nids(struct f2fs_sb_info *sbi,
4d57b86d 2460 bool sync, bool mount)
e05df3b1 2461{
e05df3b1 2462 struct f2fs_nm_info *nm_i = NM_I(sbi);
e2374015 2463 int i = 0, ret;
55008d84 2464 nid_t nid = nm_i->next_scan_nid;
e05df3b1 2465
e9cdd307
YH		 2466 if (unlikely(nid >= nm_i->max_nid))
2467 nid = 0;
2468
e2cab031
ST		 2469 if (unlikely(nid % NAT_ENTRY_PER_BLOCK))
2470 nid = NAT_BLOCK_OFFSET(nid) * NAT_ENTRY_PER_BLOCK;
2471
55008d84 2472 /* Enough entries */
9a4ffdf5 2473 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2474 return 0;
e05df3b1 2475
4d57b86d 2476 if (!sync && !f2fs_available_free_memory(sbi, FREE_NIDS))
e2374015 2477 return 0;
e05df3b1 2478
4ac91242
CY		 2479 if (!mount) {
2480 /* try to find free nids in free_nid_bitmap */
2481 scan_free_nid_bits(sbi);
2482
74986213 2483 if (nm_i->nid_cnt[FREE_NID] >= NAT_ENTRY_PER_BLOCK)
e2374015 2484 return 0;
22ad0b6a
JK		 2485 }
2486
55008d84 2487 /* readahead nat pages to be scanned */
4d57b86d 2488 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nid), FREE_NID_PAGES,
26879fb1 2489 META_NAT, true);
e05df3b1 2490
e4544b63 2491 f2fs_down_read(&nm_i->nat_tree_lock);
a5131193 2492
e05df3b1 2493 while (1) {
66e83361
YH		 2494 if (!test_bit_le(NAT_BLOCK_OFFSET(nid),
2495 nm_i->nat_block_bitmap)) {
2496 struct page *page = get_current_nat_page(sbi, nid);
e05df3b1 2497
edc55aaf
JK		 2498 if (IS_ERR(page)) {
2499 ret = PTR_ERR(page);
2500 } else {
2501 ret = scan_nat_page(sbi, page, nid);
2502 f2fs_put_page(page, 1);
2503 }
e2374015
CY		 2504
2505 if (ret) {
e4544b63 2506 f2fs_up_read(&nm_i->nat_tree_lock);
dcbb4c10 2507 f2fs_err(sbi, "NAT is corrupt, run fsck to fix it");
edc55aaf 2508 return ret;
e2374015 2509 }
66e83361 2510 }
e05df3b1
JK		 2511
2512 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK));
cfb271d4 2513 if (unlikely(nid >= nm_i->max_nid))
e05df3b1 2514 nid = 0;
55008d84 2515
a6d494b6 2516 if (++i >= FREE_NID_PAGES)
e05df3b1
JK		 2517 break;
2518 }
2519
55008d84
JK		 2520 /* go to the next free nat pages to find free nids abundantly */
2521 nm_i->next_scan_nid = nid;
e05df3b1
JK		 2522
2523 /* find free nids from current sum_pages */
2fbaa25f 2524 scan_curseg_cache(sbi);
dfc08a12 2525
e4544b63 2526 f2fs_up_read(&nm_i->nat_tree_lock);
2db2388f 2527
4d57b86d 2528 f2fs_ra_meta_pages(sbi, NAT_BLOCK_OFFSET(nm_i->next_scan_nid),
ea1a29a0 2529 nm_i->ra_nid_pages, META_NAT, false);
e2374015
CY		 2530
2531 return 0;
e05df3b1
JK		 2532}
2533
e2374015 2534int f2fs_build_free_nids(struct f2fs_sb_info *sbi, bool sync, bool mount)
2411cf5b 2535{
e2374015
CY		 2536 int ret;
2537
2411cf5b 2538 mutex_lock(&NM_I(sbi)->build_lock);
e2374015 2539 ret = __f2fs_build_free_nids(sbi, sync, mount);
2411cf5b 2540 mutex_unlock(&NM_I(sbi)->build_lock);
e2374015
CY		 2541
2542 return ret;
2411cf5b
CY		 2543}
2544
e05df3b1
JK		 2545/*
2546 * If this function returns success, caller can obtain a new nid
2547 * from second parameter of this function.
2548 * The returned nid could be used ino as well as nid when inode is created.
2549 */
4d57b86d 2550bool f2fs_alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid)
e05df3b1
JK		 2551{
2552 struct f2fs_nm_info *nm_i = NM_I(sbi);
2553 struct free_nid *i = NULL;
e05df3b1 2554retry:
c40e15a9 2555 if (time_to_inject(sbi, FAULT_ALLOC_NID))
cb78942b 2556 return false;
7fa750a1 2557
b8559dc2 2558 spin_lock(&nm_i->nid_list_lock);
e05df3b1 2559
04d47e67
CY		 2560 if (unlikely(nm_i->available_nids == 0)) {
2561 spin_unlock(&nm_i->nid_list_lock);
2562 return false;
2563 }
e05df3b1 2564
4d57b86d
CY		 2565 /* We should not use stale free nids created by f2fs_build_free_nids */
2566 if (nm_i->nid_cnt[FREE_NID] && !on_f2fs_build_free_nids(nm_i)) {
9a4ffdf5
CY		 2567 f2fs_bug_on(sbi, list_empty(&nm_i->free_nid_list));
2568 i = list_first_entry(&nm_i->free_nid_list,
b8559dc2 2569 struct free_nid, list);
55008d84 2570 *nid = i->nid;
b8559dc2 2571
a0761f63 2572 __move_free_nid(sbi, i, FREE_NID, PREALLOC_NID);
04d47e67 2573 nm_i->available_nids--;
4ac91242 2574
346fe752 2575 update_free_nid_bitmap(sbi, *nid, false, false);
4ac91242 2576
b8559dc2 2577 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2578 return true;
2579 }
b8559dc2 2580 spin_unlock(&nm_i->nid_list_lock);
55008d84
JK		 2581
2582 /* Let's scan nat pages and its caches to get free nids */
f84262b0
JK		 2583 if (!f2fs_build_free_nids(sbi, true, false))
2584 goto retry;
2585 return false;
e05df3b1
JK		 2586}
2587
0a8165d7 2588/*
4d57b86d 2589 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2590 */
4d57b86d 2591void f2fs_alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1
JK		 2592{
2593 struct f2fs_nm_info *nm_i = NM_I(sbi);
2594 struct free_nid *i;
2595
b8559dc2 2596 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2597 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2 2598 f2fs_bug_on(sbi, !i);
a0761f63 2599 __remove_free_nid(sbi, i, PREALLOC_NID);
b8559dc2 2600 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2601
2602 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2603}
2604
0a8165d7 2605/*
4d57b86d 2606 * f2fs_alloc_nid() should be called prior to this function.
e05df3b1 2607 */
4d57b86d 2608void f2fs_alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid)
e05df3b1 2609{
49952fa1
JK		 2610 struct f2fs_nm_info *nm_i = NM_I(sbi);
2611 struct free_nid *i;
cf0ee0f0 2612 bool need_free = false;
49952fa1 2613
65985d93
JK		 2614 if (!nid)
2615 return;
2616
b8559dc2 2617 spin_lock(&nm_i->nid_list_lock);
8a7ed66a 2618 i = __lookup_free_nid_list(nm_i, nid);
b8559dc2
CY		 2619 f2fs_bug_on(sbi, !i);
2620
4d57b86d 2621 if (!f2fs_available_free_memory(sbi, FREE_NIDS)) {
a0761f63 2622 __remove_free_nid(sbi, i, PREALLOC_NID);
cf0ee0f0 2623 need_free = true;
95630cba 2624 } else {
a0761f63 2625 __move_free_nid(sbi, i, PREALLOC_NID, FREE_NID);
95630cba 2626 }
04d47e67
CY		 2627
2628 nm_i->available_nids++;
2629
346fe752 2630 update_free_nid_bitmap(sbi, nid, true, false);
4ac91242 2631
b8559dc2 2632 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0
CY		 2633
2634 if (need_free)
2635 kmem_cache_free(free_nid_slab, i);
e05df3b1
JK		 2636}
2637
4d57b86d 2638int f2fs_try_to_free_nids(struct f2fs_sb_info *sbi, int nr_shrink)
31696580
CY		 2639{
2640 struct f2fs_nm_info *nm_i = NM_I(sbi);
31696580
CY		 2641 int nr = nr_shrink;
2642
9a4ffdf5 2643 if (nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
ad4edb83
JK		 2644 return 0;
2645
31696580
CY		 2646 if (!mutex_trylock(&nm_i->build_lock))
2647 return 0;
2648
042be373
CY		 2649 while (nr_shrink && nm_i->nid_cnt[FREE_NID] > MAX_FREE_NIDS) {
2650 struct free_nid *i, *next;
2651 unsigned int batch = SHRINK_NID_BATCH_SIZE;
b8559dc2 2652
042be373
CY		 2653 spin_lock(&nm_i->nid_list_lock);
2654 list_for_each_entry_safe(i, next, &nm_i->free_nid_list, list) {
2655 if (!nr_shrink || !batch ||
2656 nm_i->nid_cnt[FREE_NID] <= MAX_FREE_NIDS)
2657 break;
2658 __remove_free_nid(sbi, i, FREE_NID);
2659 kmem_cache_free(free_nid_slab, i);
2660 nr_shrink--;
2661 batch--;
2662 }
2663 spin_unlock(&nm_i->nid_list_lock);
31696580 2664 }
042be373 2665
31696580
CY		 2666 mutex_unlock(&nm_i->build_lock);
2667
2668 return nr - nr_shrink;
2669}
2670
9627a7b3 2671int f2fs_recover_inline_xattr(struct inode *inode, struct page *page)
28cdce04 2672{
28cdce04
CY		 2673 void *src_addr, *dst_addr;
2674 size_t inline_size;
2675 struct page *ipage;
2676 struct f2fs_inode *ri;
2677
4d57b86d 2678 ipage = f2fs_get_node_page(F2FS_I_SB(inode), inode->i_ino);
9627a7b3
CY		 2679 if (IS_ERR(ipage))
2680 return PTR_ERR(ipage);
28cdce04 2681
e3b4d43f 2682 ri = F2FS_INODE(page);
1eca05aa 2683 if (ri->i_inline & F2FS_INLINE_XATTR) {
96dd0251
CY		 2684 if (!f2fs_has_inline_xattr(inode)) {
2685 set_inode_flag(inode, FI_INLINE_XATTR);
2686 stat_inc_inline_xattr(inode);
2687 }
1eca05aa 2688 } else {
96dd0251
CY		 2689 if (f2fs_has_inline_xattr(inode)) {
2690 stat_dec_inline_xattr(inode);
2691 clear_inode_flag(inode, FI_INLINE_XATTR);
2692 }
e3b4d43f
JK		 2693 goto update_inode;
2694 }
2695
6afc662e
CY		 2696 dst_addr = inline_xattr_addr(inode, ipage);
2697 src_addr = inline_xattr_addr(inode, page);
28cdce04
CY		 2698 inline_size = inline_xattr_size(inode);
2699
bae0ee7a 2700 f2fs_wait_on_page_writeback(ipage, NODE, true, true);
28cdce04 2701 memcpy(dst_addr, src_addr, inline_size);
e3b4d43f 2702update_inode:
4d57b86d 2703 f2fs_update_inode(inode, ipage);
28cdce04 2704 f2fs_put_page(ipage, 1);
9627a7b3 2705 return 0;
28cdce04
CY		 2706}
2707
4d57b86d 2708int f2fs_recover_xattr_data(struct inode *inode, struct page *page)
abb2366c 2709{
4081363f 2710 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
abb2366c 2711 nid_t prev_xnid = F2FS_I(inode)->i_xattr_nid;
87905682
YH		 2712 nid_t new_xnid;
2713 struct dnode_of_data dn;
abb2366c 2714 struct node_info ni;
d260081c 2715 struct page *xpage;
7735730d 2716 int err;
abb2366c 2717
abb2366c
JK		 2718 if (!prev_xnid)
2719 goto recover_xnid;
2720
d260081c 2721 /* 1: invalidate the previous xattr nid */
a9419b63 2722 err = f2fs_get_node_info(sbi, prev_xnid, &ni, false);
7735730d
CY		 2723 if (err)
2724 return err;
2725
4d57b86d 2726 f2fs_invalidate_blocks(sbi, ni.blk_addr);
000519f2 2727 dec_valid_node_count(sbi, inode, false);
479f40c4 2728 set_node_addr(sbi, &ni, NULL_ADDR, false);
abb2366c
JK		 2729
2730recover_xnid:
d260081c 2731 /* 2: update xattr nid in inode */
4d57b86d 2732 if (!f2fs_alloc_nid(sbi, &new_xnid))
87905682
YH		 2733 return -ENOSPC;
2734
2735 set_new_dnode(&dn, inode, NULL, NULL, new_xnid);
4d57b86d 2736 xpage = f2fs_new_node_page(&dn, XATTR_NODE_OFFSET);
87905682 2737 if (IS_ERR(xpage)) {
4d57b86d 2738 f2fs_alloc_nid_failed(sbi, new_xnid);
87905682
YH		 2739 return PTR_ERR(xpage);
2740 }
2741
4d57b86d
CY		 2742 f2fs_alloc_nid_done(sbi, new_xnid);
2743 f2fs_update_inode_page(inode);
d260081c
CY		 2744
2745 /* 3: update and set xattr node page dirty */
87905682 2746 memcpy(F2FS_NODE(xpage), F2FS_NODE(page), VALID_XATTR_BLOCK_SIZE);
abb2366c 2747
d260081c
CY		 2748 set_page_dirty(xpage);
2749 f2fs_put_page(xpage, 1);
abb2366c 2750
d260081c 2751 return 0;
abb2366c
JK		 2752}
2753
4d57b86d 2754int f2fs_recover_inode_page(struct f2fs_sb_info *sbi, struct page *page)
e05df3b1 2755{
58bfaf44 2756 struct f2fs_inode *src, *dst;
e05df3b1
JK		 2757 nid_t ino = ino_of_node(page);
2758 struct node_info old_ni, new_ni;
2759 struct page *ipage;
7735730d 2760 int err;
e05df3b1 2761
a9419b63 2762 err = f2fs_get_node_info(sbi, ino, &old_ni, false);
7735730d
CY		 2763 if (err)
2764 return err;
e8271fa3
JK		 2765
2766 if (unlikely(old_ni.blk_addr != NULL_ADDR))
2767 return -EINVAL;
e8ea9b3d 2768retry:
300e129c 2769 ipage = f2fs_grab_cache_page(NODE_MAPPING(sbi), ino, false);
e8ea9b3d 2770 if (!ipage) {
4034247a 2771 memalloc_retry_wait(GFP_NOFS);
e8ea9b3d
JK		 2772 goto retry;
2773 }
e05df3b1 2774
e1c42045 2775 /* Should not use this inode from free nid list */
b8559dc2 2776 remove_free_nid(sbi, ino);
e05df3b1 2777
237c0790
JK		 2778 if (!PageUptodate(ipage))
2779 SetPageUptodate(ipage);
e05df3b1 2780 fill_node_footer(ipage, ino, ino, 0, true);
ef2a0071 2781 set_cold_node(ipage, false);
e05df3b1 2782
58bfaf44
JK		 2783 src = F2FS_INODE(page);
2784 dst = F2FS_INODE(ipage);
e05df3b1 2785
36218b81 2786 memcpy(dst, src, offsetof(struct f2fs_inode, i_ext));
58bfaf44
JK		 2787 dst->i_size = 0;
2788 dst->i_blocks = cpu_to_le64(1);
2789 dst->i_links = cpu_to_le32(1);
2790 dst->i_xattr_nid = 0;
7a2af766 2791 dst->i_inline = src->i_inline & (F2FS_INLINE_XATTR | F2FS_EXTRA_ATTR);
5c57132e 2792 if (dst->i_inline & F2FS_EXTRA_ATTR) {
7a2af766 2793 dst->i_extra_isize = src->i_extra_isize;
6afc662e 2794
7beb01f7 2795 if (f2fs_sb_has_flexible_inline_xattr(sbi) &&
6afc662e
CY		 2796 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2797 i_inline_xattr_size))
2798 dst->i_inline_xattr_size = src->i_inline_xattr_size;
2799
7beb01f7 2800 if (f2fs_sb_has_project_quota(sbi) &&
5c57132e
CY		 2801 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2802 i_projid))
2803 dst->i_projid = src->i_projid;
5cd1f387 2804
7beb01f7 2805 if (f2fs_sb_has_inode_crtime(sbi) &&
5cd1f387
CY		 2806 F2FS_FITS_IN_INODE(src, le16_to_cpu(src->i_extra_isize),
2807 i_crtime_nsec)) {
2808 dst->i_crtime = src->i_crtime;
2809 dst->i_crtime_nsec = src->i_crtime_nsec;
2810 }
5c57132e 2811 }
e05df3b1
JK		 2812
2813 new_ni = old_ni;
2814 new_ni.ino = ino;
2815
0abd675e 2816 if (unlikely(inc_valid_node_count(sbi, NULL, true)))
65e5cd0a 2817 WARN_ON(1);
479f40c4 2818 set_node_addr(sbi, &new_ni, NEW_ADDR, false);
e05df3b1 2819 inc_valid_inode_count(sbi);
617deb8c 2820 set_page_dirty(ipage);
e05df3b1
JK		 2821 f2fs_put_page(ipage, 1);
2822 return 0;
2823}
2824
7735730d 2825int f2fs_restore_node_summary(struct f2fs_sb_info *sbi,
e05df3b1
JK		 2826 unsigned int segno, struct f2fs_summary_block *sum)
2827{
2828 struct f2fs_node *rn;
2829 struct f2fs_summary *sum_entry;
e05df3b1 2830 block_t addr;
9ecf4b80 2831 int i, idx, last_offset, nrpages;
e05df3b1
JK		 2832
2833 /* scan the node segment */
2834 last_offset = sbi->blocks_per_seg;
2835 addr = START_BLOCK(sbi, segno);
2836 sum_entry = &sum->entries[0];
2837
9ecf4b80 2838 for (i = 0; i < last_offset; i += nrpages, addr += nrpages) {
5f7136db 2839 nrpages = bio_max_segs(last_offset - i);
393ff91f 2840
e1c42045 2841 /* readahead node pages */
4d57b86d 2842 f2fs_ra_meta_pages(sbi, addr, nrpages, META_POR, true);
e05df3b1 2843
9ecf4b80 2844 for (idx = addr; idx < addr + nrpages; idx++) {
4d57b86d 2845 struct page *page = f2fs_get_tmp_page(sbi, idx);
9af0ff1c 2846
7735730d
CY		 2847 if (IS_ERR(page))
2848 return PTR_ERR(page);
2849
9ecf4b80
CY		 2850 rn = F2FS_NODE(page);
2851 sum_entry->nid = rn->footer.nid;
2852 sum_entry->version = 0;
2853 sum_entry->ofs_in_node = 0;
2854 sum_entry++;
2855 f2fs_put_page(page, 1);
9af0ff1c 2856 }
bac4eef6 2857
9ecf4b80 2858 invalidate_mapping_pages(META_MAPPING(sbi), addr,
bac4eef6 2859 addr + nrpages);
e05df3b1 2860 }
7735730d 2861 return 0;
e05df3b1
JK		 2862}
2863
aec71382 2864static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
e05df3b1
JK		 2865{
2866 struct f2fs_nm_info *nm_i = NM_I(sbi);
2867 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2868 struct f2fs_journal *journal = curseg->journal;
e05df3b1
JK		 2869 int i;
2870
b7ad7512 2871 down_write(&curseg->journal_rwsem);
dfc08a12 2872 for (i = 0; i < nats_in_cursum(journal); i++) {
e05df3b1
JK		 2873 struct nat_entry *ne;
2874 struct f2fs_nat_entry raw_ne;
dfc08a12 2875 nid_t nid = le32_to_cpu(nid_in_journal(journal, i));
e05df3b1 2876
b862676e
CY		 2877 if (f2fs_check_nid_range(sbi, nid))
2878 continue;
2879
dfc08a12 2880 raw_ne = nat_in_journal(journal, i);
9be32d72 2881
e05df3b1 2882 ne = __lookup_nat_cache(nm_i, nid);
e05df3b1 2883 if (!ne) {
32410577 2884 ne = __alloc_nat_entry(sbi, nid, true);
12f9ef37 2885 __init_nat_entry(nm_i, ne, &raw_ne, true);
e05df3b1 2886 }
04d47e67
CY		 2887
2888 /*
2889 * if a free nat in journal has not been used after last
2890 * checkpoint, we should remove it from available nids,
2891 * since later we will add it again.
2892 */
2893 if (!get_nat_flag(ne, IS_DIRTY) &&
2894 le32_to_cpu(raw_ne.block_addr) == NULL_ADDR) {
2895 spin_lock(&nm_i->nid_list_lock);
2896 nm_i->available_nids--;
2897 spin_unlock(&nm_i->nid_list_lock);
2898 }
2899
e05df3b1 2900 __set_nat_cache_dirty(nm_i, ne);
e05df3b1 2901 }
dfc08a12 2902 update_nats_in_cursum(journal, -i);
b7ad7512 2903 up_write(&curseg->journal_rwsem);
e05df3b1
JK		 2904}
2905
309cc2b6
JK		 2906static void __adjust_nat_entry_set(struct nat_entry_set *nes,
2907 struct list_head *head, int max)
e05df3b1 2908{
309cc2b6 2909 struct nat_entry_set *cur;
e05df3b1 2910
309cc2b6
JK		 2911 if (nes->entry_cnt >= max)
2912 goto add_out;
e05df3b1 2913
309cc2b6
JK		 2914 list_for_each_entry(cur, head, set_list) {
2915 if (cur->entry_cnt >= nes->entry_cnt) {
2916 list_add(&nes->set_list, cur->set_list.prev);
2917 return;
2918 }
aec71382 2919 }
309cc2b6
JK		 2920add_out:
2921 list_add_tail(&nes->set_list, head);
2922}
e05df3b1 2923
94c821fb
CY		 2924static void __update_nat_bits(struct f2fs_nm_info *nm_i, unsigned int nat_ofs,
2925 unsigned int valid)
2926{
2927 if (valid == 0) {
2928 __set_bit_le(nat_ofs, nm_i->empty_nat_bits);
2929 __clear_bit_le(nat_ofs, nm_i->full_nat_bits);
2930 return;
2931 }
2932
2933 __clear_bit_le(nat_ofs, nm_i->empty_nat_bits);
2934 if (valid == NAT_ENTRY_PER_BLOCK)
2935 __set_bit_le(nat_ofs, nm_i->full_nat_bits);
2936 else
2937 __clear_bit_le(nat_ofs, nm_i->full_nat_bits);
2938}
2939
2940static void update_nat_bits(struct f2fs_sb_info *sbi, nid_t start_nid,
22ad0b6a
JK		 2941 struct page *page)
2942{
2943 struct f2fs_nm_info *nm_i = NM_I(sbi);
2944 unsigned int nat_index = start_nid / NAT_ENTRY_PER_BLOCK;
2945 struct f2fs_nat_block *nat_blk = page_address(page);
2946 int valid = 0;
37a0ab2a 2947 int i = 0;
22ad0b6a 2948
94c821fb 2949 if (!is_set_ckpt_flags(sbi, CP_NAT_BITS_FLAG))
22ad0b6a
JK		 2950 return;
2951
37a0ab2a
FL		 2952 if (nat_index == 0) {
2953 valid = 1;
2954 i = 1;
2955 }
2956 for (; i < NAT_ENTRY_PER_BLOCK; i++) {
36af5f40 2957 if (le32_to_cpu(nat_blk->entries[i].block_addr) != NULL_ADDR)
22ad0b6a
JK		 2958 valid++;
2959 }
94c821fb
CY		 2960
2961 __update_nat_bits(nm_i, nat_index, valid);
2962}
2963
2964void f2fs_enable_nat_bits(struct f2fs_sb_info *sbi)
2965{
2966 struct f2fs_nm_info *nm_i = NM_I(sbi);
2967 unsigned int nat_ofs;
2968
e4544b63 2969 f2fs_down_read(&nm_i->nat_tree_lock);
94c821fb
CY		 2970
2971 for (nat_ofs = 0; nat_ofs < nm_i->nat_blocks; nat_ofs++) {
2972 unsigned int valid = 0, nid_ofs = 0;
2973
2974 /* handle nid zero due to it should never be used */
2975 if (unlikely(nat_ofs == 0)) {
2976 valid = 1;
2977 nid_ofs = 1;
2978 }
2979
2980 for (; nid_ofs < NAT_ENTRY_PER_BLOCK; nid_ofs++) {
2981 if (!test_bit_le(nid_ofs,
2982 nm_i->free_nid_bitmap[nat_ofs]))
2983 valid++;
2984 }
2985
2986 __update_nat_bits(nm_i, nat_ofs, valid);
22ad0b6a
JK		 2987 }
2988
e4544b63 2989 f2fs_up_read(&nm_i->nat_tree_lock);
22ad0b6a
JK		 2990}
2991
edc55aaf 2992static int __flush_nat_entry_set(struct f2fs_sb_info *sbi,
22ad0b6a 2993 struct nat_entry_set *set, struct cp_control *cpc)
309cc2b6
JK		 2994{
2995 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 2996 struct f2fs_journal *journal = curseg->journal;
309cc2b6
JK		 2997 nid_t start_nid = set->set * NAT_ENTRY_PER_BLOCK;
2998 bool to_journal = true;
2999 struct f2fs_nat_block *nat_blk;
3000 struct nat_entry *ne, *cur;
3001 struct page *page = NULL;
e05df3b1 3002
aec71382
CY		 3003 /*
3004 * there are two steps to flush nat entries:
3005 * #1, flush nat entries to journal in current hot data summary block.
3006 * #2, flush nat entries to nat page.
3007 */
94c821fb 3008 if ((cpc->reason & CP_UMOUNT) ||
22ad0b6a 3009 !__has_cursum_space(journal, set->entry_cnt, NAT_JOURNAL))
309cc2b6
JK		 3010 to_journal = false;
3011
3012 if (to_journal) {
b7ad7512 3013 down_write(&curseg->journal_rwsem);
309cc2b6
JK		 3014 } else {
3015 page = get_next_nat_page(sbi, start_nid);
edc55aaf
JK		 3016 if (IS_ERR(page))
3017 return PTR_ERR(page);
3018
309cc2b6
JK		 3019 nat_blk = page_address(page);
3020 f2fs_bug_on(sbi, !nat_blk);
3021 }
aec71382 3022
309cc2b6
JK		 3023 /* flush dirty nats in nat entry set */
3024 list_for_each_entry_safe(ne, cur, &set->entry_list, list) {
3025 struct f2fs_nat_entry *raw_ne;
3026 nid_t nid = nat_get_nid(ne);
3027 int offset;
3028
febeca6d 3029 f2fs_bug_on(sbi, nat_get_blkaddr(ne) == NEW_ADDR);
aec71382
CY		 3030
3031 if (to_journal) {
4d57b86d 3032 offset = f2fs_lookup_journal_in_cursum(journal,
309cc2b6
JK		 3033 NAT_JOURNAL, nid, 1);
3034 f2fs_bug_on(sbi, offset < 0);
dfc08a12
CY		 3035 raw_ne = &nat_in_journal(journal, offset);
3036 nid_in_journal(journal, offset) = cpu_to_le32(nid);
aec71382 3037 } else {
309cc2b6 3038 raw_ne = &nat_blk->entries[nid - start_nid];
e05df3b1 3039 }
309cc2b6 3040 raw_nat_from_node_info(raw_ne, &ne->ni);
309cc2b6 3041 nat_reset_flag(ne);
0b28b71e 3042 __clear_nat_cache_dirty(NM_I(sbi), set, ne);
04d47e67 3043 if (nat_get_blkaddr(ne) == NULL_ADDR) {
5921aaa1 3044 add_free_nid(sbi, nid, false, true);
4ac91242
CY		 3045 } else {
3046 spin_lock(&NM_I(sbi)->nid_list_lock);
346fe752 3047 update_free_nid_bitmap(sbi, nid, false, false);
04d47e67
CY		 3048 spin_unlock(&NM_I(sbi)->nid_list_lock);
3049 }
309cc2b6 3050 }
e05df3b1 3051
22ad0b6a 3052 if (to_journal) {
b7ad7512 3053 up_write(&curseg->journal_rwsem);
22ad0b6a 3054 } else {
94c821fb 3055 update_nat_bits(sbi, start_nid, page);
309cc2b6 3056 f2fs_put_page(page, 1);
22ad0b6a 3057 }
aec71382 3058
59c9081b
YH		 3059 /* Allow dirty nats by node block allocation in write_begin */
3060 if (!set->entry_cnt) {
3061 radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set);
3062 kmem_cache_free(nat_entry_set_slab, set);
3063 }
edc55aaf 3064 return 0;
309cc2b6 3065}
aec71382 3066
309cc2b6
JK		 3067/*
3068 * This function is called during the checkpointing process.
3069 */
edc55aaf 3070int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
309cc2b6
JK		 3071{
3072 struct f2fs_nm_info *nm_i = NM_I(sbi);
3073 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
b7ad7512 3074 struct f2fs_journal *journal = curseg->journal;
c31e4961 3075 struct nat_entry_set *setvec[NAT_VEC_SIZE];
309cc2b6
JK		 3076 struct nat_entry_set *set, *tmp;
3077 unsigned int found;
3078 nid_t set_idx = 0;
3079 LIST_HEAD(sets);
edc55aaf 3080 int err = 0;
309cc2b6 3081
a95ba66a
JK		 3082 /*
3083 * during unmount, let's flush nat_bits before checking
3084 * nat_cnt[DIRTY_NAT].
3085 */
94c821fb 3086 if (cpc->reason & CP_UMOUNT) {
e4544b63 3087 f2fs_down_write(&nm_i->nat_tree_lock);
7f2ecdd8 3088 remove_nats_in_journal(sbi);
e4544b63 3089 f2fs_up_write(&nm_i->nat_tree_lock);
7f2ecdd8
JK		 3090 }
3091
a95ba66a 3092 if (!nm_i->nat_cnt[DIRTY_NAT])
edc55aaf 3093 return 0;
a5131193 3094
e4544b63 3095 f2fs_down_write(&nm_i->nat_tree_lock);
a5131193 3096
309cc2b6
JK		 3097 /*
3098 * if there are no enough space in journal to store dirty nat
3099 * entries, remove all entries from journal and merge them
3100 * into nat entry set.
3101 */
94c821fb 3102 if (cpc->reason & CP_UMOUNT ||
a95ba66a
JK		 3103 !__has_cursum_space(journal,
3104 nm_i->nat_cnt[DIRTY_NAT], NAT_JOURNAL))
309cc2b6
JK		 3105 remove_nats_in_journal(sbi);
3106
309cc2b6 3107 while ((found = __gang_lookup_nat_set(nm_i,
c31e4961 3108 set_idx, NAT_VEC_SIZE, setvec))) {
309cc2b6 3109 unsigned idx;
5f029c04 3110
309cc2b6
JK		 3111 set_idx = setvec[found - 1]->set + 1;
3112 for (idx = 0; idx < found; idx++)
3113 __adjust_nat_entry_set(setvec[idx], &sets,
dfc08a12 3114 MAX_NAT_JENTRIES(journal));
e05df3b1 3115 }
aec71382 3116
309cc2b6 3117 /* flush dirty nats in nat entry set */
edc55aaf
JK		 3118 list_for_each_entry_safe(set, tmp, &sets, set_list) {
3119 err = __flush_nat_entry_set(sbi, set, cpc);
3120 if (err)
3121 break;
3122 }
309cc2b6 3123
e4544b63 3124 f2fs_up_write(&nm_i->nat_tree_lock);
59c9081b 3125 /* Allow dirty nats by node block allocation in write_begin */
edc55aaf
JK		 3126
3127 return err;
e05df3b1
JK		 3128}
3129
22ad0b6a
JK		 3130static int __get_nat_bitmaps(struct f2fs_sb_info *sbi)
3131{
3132 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3133 struct f2fs_nm_info *nm_i = NM_I(sbi);
3134 unsigned int nat_bits_bytes = nm_i->nat_blocks / BITS_PER_BYTE;
3135 unsigned int i;
3136 __u64 cp_ver = cur_cp_version(ckpt);
22ad0b6a
JK		 3137 block_t nat_bits_addr;
3138
df033caf 3139 nm_i->nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
0b6d4ca0 3140 nm_i->nat_bits = f2fs_kvzalloc(sbi,
acbf054d 3141 nm_i->nat_bits_blocks << F2FS_BLKSIZE_BITS, GFP_KERNEL);
22ad0b6a
JK		 3142 if (!nm_i->nat_bits)
3143 return -ENOMEM;
3144
94c821fb
CY		 3145 nm_i->full_nat_bits = nm_i->nat_bits + 8;
3146 nm_i->empty_nat_bits = nm_i->full_nat_bits + nat_bits_bytes;
3147
3148 if (!is_set_ckpt_flags(sbi, CP_NAT_BITS_FLAG))
3149 return 0;
3150
22ad0b6a
JK		 3151 nat_bits_addr = __start_cp_addr(sbi) + sbi->blocks_per_seg -
3152 nm_i->nat_bits_blocks;
3153 for (i = 0; i < nm_i->nat_bits_blocks; i++) {
7735730d
CY		 3154 struct page *page;
3155
3156 page = f2fs_get_meta_page(sbi, nat_bits_addr++);
3b30eb19 3157 if (IS_ERR(page))
7735730d 3158 return PTR_ERR(page);
22ad0b6a
JK		 3159
3160 memcpy(nm_i->nat_bits + (i << F2FS_BLKSIZE_BITS),
3161 page_address(page), F2FS_BLKSIZE);
3162 f2fs_put_page(page, 1);
3163 }
3164
ced2c7ea 3165 cp_ver |= (cur_cp_crc(ckpt) << 32);
22ad0b6a 3166 if (cpu_to_le64(cp_ver) != *(__le64 *)nm_i->nat_bits) {
94c821fb
CY		 3167 clear_ckpt_flags(sbi, CP_NAT_BITS_FLAG);
3168 f2fs_notice(sbi, "Disable nat_bits due to incorrect cp_ver (%llu, %llu)",
3169 cp_ver, le64_to_cpu(*(__le64 *)nm_i->nat_bits));
22ad0b6a
JK		 3170 return 0;
3171 }
3172
dcbb4c10 3173 f2fs_notice(sbi, "Found nat_bits in checkpoint");
22ad0b6a
JK		 3174 return 0;
3175}
3176
bd80a4b9 3177static inline void load_free_nid_bitmap(struct f2fs_sb_info *sbi)
7041d5d2
CY		 3178{
3179 struct f2fs_nm_info *nm_i = NM_I(sbi);
3180 unsigned int i = 0;
3181 nid_t nid, last_nid;
3182
94c821fb 3183 if (!is_set_ckpt_flags(sbi, CP_NAT_BITS_FLAG))
7041d5d2
CY		 3184 return;
3185
3186 for (i = 0; i < nm_i->nat_blocks; i++) {
3187 i = find_next_bit_le(nm_i->empty_nat_bits, nm_i->nat_blocks, i);
3188 if (i >= nm_i->nat_blocks)
3189 break;
3190
3191 __set_bit_le(i, nm_i->nat_block_bitmap);
3192
3193 nid = i * NAT_ENTRY_PER_BLOCK;
f6986ede 3194 last_nid = nid + NAT_ENTRY_PER_BLOCK;
7041d5d2 3195
346fe752 3196 spin_lock(&NM_I(sbi)->nid_list_lock);
7041d5d2 3197 for (; nid < last_nid; nid++)
346fe752
CY		 3198 update_free_nid_bitmap(sbi, nid, true, true);
3199 spin_unlock(&NM_I(sbi)->nid_list_lock);
7041d5d2
CY		 3200 }
3201
3202 for (i = 0; i < nm_i->nat_blocks; i++) {
3203 i = find_next_bit_le(nm_i->full_nat_bits, nm_i->nat_blocks, i);
3204 if (i >= nm_i->nat_blocks)
3205 break;
3206
3207 __set_bit_le(i, nm_i->nat_block_bitmap);
3208 }
3209}
3210
e05df3b1
JK		 3211static int init_node_manager(struct f2fs_sb_info *sbi)
3212{
3213 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi);
3214 struct f2fs_nm_info *nm_i = NM_I(sbi);
3215 unsigned char *version_bitmap;
22ad0b6a
JK		 3216 unsigned int nat_segs;
3217 int err;
e05df3b1
JK		 3218
3219 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr);
3220
3221 /* segment_count_nat includes pair segment so divide to 2. */
3222 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1;
22ad0b6a
JK		 3223 nm_i->nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg);
3224 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
7ee0eeab 3225
b63da15e 3226 /* not used nids: 0, node, meta, (and root counted as valid node) */
04d47e67 3227 nm_i->available_nids = nm_i->max_nid - sbi->total_valid_node_count -
27cae0bc 3228 F2FS_RESERVED_NODE_NUM;
9a4ffdf5
CY		 3229 nm_i->nid_cnt[FREE_NID] = 0;
3230 nm_i->nid_cnt[PREALLOC_NID] = 0;
cdfc41c1 3231 nm_i->ram_thresh = DEF_RAM_THRESHOLD;
ea1a29a0 3232 nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
2304cb0c 3233 nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
47c8ebcc 3234 nm_i->max_rf_node_blocks = DEF_RF_NODE_BLOCKS;
e05df3b1 3235
8a7ed66a 3236 INIT_RADIX_TREE(&nm_i->free_nid_root, GFP_ATOMIC);
9a4ffdf5 3237 INIT_LIST_HEAD(&nm_i->free_nid_list);
769ec6e5
JK		 3238 INIT_RADIX_TREE(&nm_i->nat_root, GFP_NOIO);
3239 INIT_RADIX_TREE(&nm_i->nat_set_root, GFP_NOIO);
e05df3b1 3240 INIT_LIST_HEAD(&nm_i->nat_entries);
22969158 3241 spin_lock_init(&nm_i->nat_list_lock);
e05df3b1
JK		 3242
3243 mutex_init(&nm_i->build_lock);
b8559dc2 3244 spin_lock_init(&nm_i->nid_list_lock);
e4544b63 3245 init_f2fs_rwsem(&nm_i->nat_tree_lock);
e05df3b1 3246
e05df3b1 3247 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid);
79b5793b 3248 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
e05df3b1 3249 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
79b5793b
AG		 3250 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size,
3251 GFP_KERNEL);
3252 if (!nm_i->nat_bitmap)
3253 return -ENOMEM;
599a09b2 3254
22ad0b6a
JK		 3255 err = __get_nat_bitmaps(sbi);
3256 if (err)
3257 return err;
3258
599a09b2
CY		 3259#ifdef CONFIG_F2FS_CHECK_FS
3260 nm_i->nat_bitmap_mir = kmemdup(version_bitmap, nm_i->bitmap_size,
3261 GFP_KERNEL);
3262 if (!nm_i->nat_bitmap_mir)
3263 return -ENOMEM;
3264#endif
3265
e05df3b1
JK		 3266 return 0;
3267}
3268
9f7e4a2c 3269static int init_free_nid_cache(struct f2fs_sb_info *sbi)
4ac91242
CY		 3270{
3271 struct f2fs_nm_info *nm_i = NM_I(sbi);
bb1105e4 3272 int i;
4ac91242 3273
026f0507 3274 nm_i->free_nid_bitmap =
0b6d4ca0
EB		 3275 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned char *),
3276 nm_i->nat_blocks),
3277 GFP_KERNEL);
4ac91242
CY		 3278 if (!nm_i->free_nid_bitmap)
3279 return -ENOMEM;
3280
bb1105e4
JK		 3281 for (i = 0; i < nm_i->nat_blocks; i++) {
3282 nm_i->free_nid_bitmap[i] = f2fs_kvzalloc(sbi,
e15d54d5 3283 f2fs_bitmap_size(NAT_ENTRY_PER_BLOCK), GFP_KERNEL);
68c43a23 3284 if (!nm_i->free_nid_bitmap[i])
bb1105e4
JK		 3285 return -ENOMEM;
3286 }
3287
628b3d14 3288 nm_i->nat_block_bitmap = f2fs_kvzalloc(sbi, nm_i->nat_blocks / 8,
4ac91242
CY		 3289 GFP_KERNEL);
3290 if (!nm_i->nat_block_bitmap)
3291 return -ENOMEM;
586d1492 3292
9d2a789c
KC		 3293 nm_i->free_nid_count =
3294 f2fs_kvzalloc(sbi, array_size(sizeof(unsigned short),
3295 nm_i->nat_blocks),
3296 GFP_KERNEL);
586d1492
CY		 3297 if (!nm_i->free_nid_count)
3298 return -ENOMEM;
4ac91242
CY		 3299 return 0;
3300}
3301
4d57b86d 3302int f2fs_build_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK		 3303{
3304 int err;
3305
acbf054d
CY		 3306 sbi->nm_info = f2fs_kzalloc(sbi, sizeof(struct f2fs_nm_info),
3307 GFP_KERNEL);
e05df3b1
JK		 3308 if (!sbi->nm_info)
3309 return -ENOMEM;
3310
3311 err = init_node_manager(sbi);
3312 if (err)
3313 return err;
3314
4ac91242
CY		 3315 err = init_free_nid_cache(sbi);
3316 if (err)
3317 return err;
3318
7041d5d2
CY		 3319 /* load free nid status from nat_bits table */
3320 load_free_nid_bitmap(sbi);
3321
e2374015 3322 return f2fs_build_free_nids(sbi, true, true);
e05df3b1
JK		 3323}
3324
4d57b86d 3325void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi)
e05df3b1
JK		 3326{
3327 struct f2fs_nm_info *nm_i = NM_I(sbi);
3328 struct free_nid *i, *next_i;
c31e4961
CY		 3329 void *vec[NAT_VEC_SIZE];
3330 struct nat_entry **natvec = (struct nat_entry **)vec;
3331 struct nat_entry_set **setvec = (struct nat_entry_set **)vec;
e05df3b1
JK		 3332 nid_t nid = 0;
3333 unsigned int found;
3334
3335 if (!nm_i)
3336 return;
3337
3338 /* destroy free nid list */
b8559dc2 3339 spin_lock(&nm_i->nid_list_lock);
9a4ffdf5 3340 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) {
a0761f63 3341 __remove_free_nid(sbi, i, FREE_NID);
b8559dc2 3342 spin_unlock(&nm_i->nid_list_lock);
cf0ee0f0 3343 kmem_cache_free(free_nid_slab, i);
b8559dc2 3344 spin_lock(&nm_i->nid_list_lock);
e05df3b1 3345 }
9a4ffdf5
CY		 3346 f2fs_bug_on(sbi, nm_i->nid_cnt[FREE_NID]);
3347 f2fs_bug_on(sbi, nm_i->nid_cnt[PREALLOC_NID]);
3348 f2fs_bug_on(sbi, !list_empty(&nm_i->free_nid_list));
b8559dc2 3349 spin_unlock(&nm_i->nid_list_lock);
e05df3b1
JK		 3350
3351 /* destroy nat cache */
e4544b63 3352 f2fs_down_write(&nm_i->nat_tree_lock);
e05df3b1 3353 while ((found = __gang_lookup_nat_cache(nm_i,
c31e4961 3354 nid, NAT_VEC_SIZE, natvec))) {
e05df3b1 3355 unsigned idx;
7aed0d45 3356
b6ce391e 3357 nid = nat_get_nid(natvec[found - 1]) + 1;
22969158
CY		 3358 for (idx = 0; idx < found; idx++) {
3359 spin_lock(&nm_i->nat_list_lock);
3360 list_del(&natvec[idx]->list);
3361 spin_unlock(&nm_i->nat_list_lock);
3362
b6ce391e 3363 __del_from_nat_cache(nm_i, natvec[idx]);
22969158 3364 }
e05df3b1 3365 }
a95ba66a 3366 f2fs_bug_on(sbi, nm_i->nat_cnt[TOTAL_NAT]);
7aed0d45
JK		 3367
3368 /* destroy nat set cache */
3369 nid = 0;
c31e4961 3370 memset(vec, 0, sizeof(void *) * NAT_VEC_SIZE);
7aed0d45 3371 while ((found = __gang_lookup_nat_set(nm_i,
c31e4961 3372 nid, NAT_VEC_SIZE, setvec))) {
7aed0d45
JK		 3373 unsigned idx;
3374
3375 nid = setvec[found - 1]->set + 1;
3376 for (idx = 0; idx < found; idx++) {
3377 /* entry_cnt is not zero, when cp_error was occurred */
3378 f2fs_bug_on(sbi, !list_empty(&setvec[idx]->entry_list));
3379 radix_tree_delete(&nm_i->nat_set_root, setvec[idx]->set);
3380 kmem_cache_free(nat_entry_set_slab, setvec[idx]);
3381 }
3382 }
e4544b63 3383 f2fs_up_write(&nm_i->nat_tree_lock);
e05df3b1 3384
4ac91242 3385 kvfree(nm_i->nat_block_bitmap);
bb1105e4
JK		 3386 if (nm_i->free_nid_bitmap) {
3387 int i;
3388
3389 for (i = 0; i < nm_i->nat_blocks; i++)
3390 kvfree(nm_i->free_nid_bitmap[i]);
5222595d 3391 kvfree(nm_i->free_nid_bitmap);
bb1105e4 3392 }
586d1492 3393 kvfree(nm_i->free_nid_count);
4ac91242 3394
5222595d
JK		 3395 kvfree(nm_i->nat_bitmap);
3396 kvfree(nm_i->nat_bits);
599a09b2 3397#ifdef CONFIG_F2FS_CHECK_FS
5222595d 3398 kvfree(nm_i->nat_bitmap_mir);
599a09b2 3399#endif
e05df3b1 3400 sbi->nm_info = NULL;
c8eb7024 3401 kfree(nm_i);
e05df3b1
JK		 3402}
3403
4d57b86d 3404int __init f2fs_create_node_manager_caches(void)
e05df3b1 3405{
98510003 3406 nat_entry_slab = f2fs_kmem_cache_create("f2fs_nat_entry",
e8512d2e 3407 sizeof(struct nat_entry));
e05df3b1 3408 if (!nat_entry_slab)
aec71382 3409 goto fail;
e05df3b1 3410
98510003 3411 free_nid_slab = f2fs_kmem_cache_create("f2fs_free_nid",
e8512d2e 3412 sizeof(struct free_nid));
aec71382 3413 if (!free_nid_slab)
ce3e6d25 3414 goto destroy_nat_entry;
aec71382 3415
98510003 3416 nat_entry_set_slab = f2fs_kmem_cache_create("f2fs_nat_entry_set",
aec71382
CY		 3417 sizeof(struct nat_entry_set));
3418 if (!nat_entry_set_slab)
ce3e6d25 3419 goto destroy_free_nid;
50fa53ec 3420
98510003 3421 fsync_node_entry_slab = f2fs_kmem_cache_create("f2fs_fsync_node_entry",
50fa53ec
CY		 3422 sizeof(struct fsync_node_entry));
3423 if (!fsync_node_entry_slab)
3424 goto destroy_nat_entry_set;
e05df3b1 3425 return 0;
aec71382 3426
50fa53ec
CY		 3427destroy_nat_entry_set:
3428 kmem_cache_destroy(nat_entry_set_slab);
ce3e6d25 3429destroy_free_nid:
aec71382 3430 kmem_cache_destroy(free_nid_slab);
ce3e6d25 3431destroy_nat_entry:
aec71382
CY		 3432 kmem_cache_destroy(nat_entry_slab);
3433fail:
3434 return -ENOMEM;
e05df3b1
JK		 3435}
3436
4d57b86d 3437void f2fs_destroy_node_manager_caches(void)
e05df3b1 3438{
50fa53ec 3439 kmem_cache_destroy(fsync_node_entry_slab);
aec71382 3440 kmem_cache_destroy(nat_entry_set_slab);
e05df3b1
JK		 3441 kmem_cache_destroy(free_nid_slab);
3442 kmem_cache_destroy(nat_entry_slab);
3443}
Linux 6.x block layer and io_uring tree(s)