projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
cifs: avoid re-lookups in dfs_cache_find()
[linux-block.git] / fs / f2fs / checkpoint.c
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
127e670a
JK		 3 * fs/f2fs/checkpoint.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
127e670a
JK		 7 */
8#include <linux/fs.h>
9#include <linux/bio.h>
10#include <linux/mpage.h>
11#include <linux/writeback.h>
12#include <linux/blkdev.h>
13#include <linux/f2fs_fs.h>
14#include <linux/pagevec.h>
15#include <linux/swap.h>
261eeb9c 16#include <linux/kthread.h>
127e670a
JK		 17
18#include "f2fs.h"
19#include "node.h"
20#include "segment.h"
52118743 21#include "iostat.h"
2af4bd6c 22#include <trace/events/f2fs.h>
127e670a 23
261eeb9c
DJ		 24#define DEFAULT_CHECKPOINT_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
25
6451e041 26static struct kmem_cache *ino_entry_slab;
4d57b86d 27struct kmem_cache *f2fs_inode_entry_slab;
127e670a 28
a9cfee0e
CY		 29void f2fs_stop_checkpoint(struct f2fs_sb_info *sbi, bool end_io,
30 unsigned char reason)
38f91ca8 31{
d494500a 32 f2fs_build_fault_attr(sbi, 0, 0);
aaec2b1d 33 set_ckpt_flags(sbi, CP_ERROR_FLAG);
a9cfee0e 34 if (!end_io) {
b9109b0e 35 f2fs_flush_merged_writes(sbi);
a9cfee0e
CY		 36
37 f2fs_handle_stop(sbi, reason);
38 }
38f91ca8
JK		 39}
40
0a8165d7 41/*
127e670a
JK		 42 * We guarantee no failure on the returned page.
43 */
4d57b86d 44struct page *f2fs_grab_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
127e670a 45{
9df27d98 46 struct address_space *mapping = META_MAPPING(sbi);
beb78181 47 struct page *page;
127e670a 48repeat:
300e129c 49 page = f2fs_grab_cache_page(mapping, index, false);
127e670a
JK		 50 if (!page) {
51 cond_resched();
52 goto repeat;
53 }
bae0ee7a 54 f2fs_wait_on_page_writeback(page, META, true, true);
237c0790
JK		 55 if (!PageUptodate(page))
56 SetPageUptodate(page);
127e670a
JK		 57 return page;
58}
59
2b947003
CY		 60static struct page *__get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index,
61 bool is_meta)
127e670a 62{
9df27d98 63 struct address_space *mapping = META_MAPPING(sbi);
127e670a 64 struct page *page;
cf04e8eb 65 struct f2fs_io_info fio = {
05ca3632 66 .sbi = sbi,
cf04e8eb 67 .type = META,
04d328de 68 .op = REQ_OP_READ,
70fd7614 69 .op_flags = REQ_META | REQ_PRIO,
7a9d7548
CY		 70 .old_blkaddr = index,
71 .new_blkaddr = index,
4375a336 72 .encrypted_page = NULL,
6dc3a126 73 .is_por = !is_meta,
cf04e8eb 74 };
7735730d 75 int err;
2b947003
CY		 76
77 if (unlikely(!is_meta))
04d328de 78 fio.op_flags &= ~REQ_META;
127e670a 79repeat:
300e129c 80 page = f2fs_grab_cache_page(mapping, index, false);
127e670a
JK		 81 if (!page) {
82 cond_resched();
83 goto repeat;
84 }
393ff91f
JK		 85 if (PageUptodate(page))
86 goto out;
87
05ca3632
JK		 88 fio.page = page;
89
7735730d
CY		 90 err = f2fs_submit_page_bio(&fio);
91 if (err) {
92 f2fs_put_page(page, 1);
93 return ERR_PTR(err);
86531d6b 94 }
127e670a 95
34a23525 96 f2fs_update_iostat(sbi, NULL, FS_META_READ_IO, F2FS_BLKSIZE);
8b83ac81 97
393ff91f 98 lock_page(page);
6bacf52f 99 if (unlikely(page->mapping != mapping)) {
afcb7ca0
JK		 100 f2fs_put_page(page, 1);
101 goto repeat;
102 }
f3f338ca 103
81114baa 104 if (unlikely(!PageUptodate(page))) {
a7b8618a 105 f2fs_handle_page_eio(sbi, page->index, META);
7735730d
CY		 106 f2fs_put_page(page, 1);
107 return ERR_PTR(-EIO);
81114baa 108 }
393ff91f 109out:
127e670a
JK		 110 return page;
111}
112
4d57b86d 113struct page *f2fs_get_meta_page(struct f2fs_sb_info *sbi, pgoff_t index)
2b947003
CY		 114{
115 return __get_meta_page(sbi, index, true);
116}
117
86f33603 118struct page *f2fs_get_meta_page_retry(struct f2fs_sb_info *sbi, pgoff_t index)
7735730d
CY		 119{
120 struct page *page;
121 int count = 0;
122
123retry:
124 page = __get_meta_page(sbi, index, true);
125 if (IS_ERR(page)) {
126 if (PTR_ERR(page) == -EIO &&
127 ++count <= DEFAULT_RETRY_IO_COUNT)
128 goto retry;
a9cfee0e 129 f2fs_stop_checkpoint(sbi, false, STOP_CP_REASON_META_PAGE);
7735730d 130 }
7735730d
CY		 131 return page;
132}
133
2b947003 134/* for POR only */
4d57b86d 135struct page *f2fs_get_tmp_page(struct f2fs_sb_info *sbi, pgoff_t index)
2b947003
CY		 136{
137 return __get_meta_page(sbi, index, false);
138}
139
93770ab7
CY		 140static bool __is_bitmap_valid(struct f2fs_sb_info *sbi, block_t blkaddr,
141 int type)
142{
143 struct seg_entry *se;
144 unsigned int segno, offset;
145 bool exist;
146
0ef4ca04 147 if (type == DATA_GENERIC)
93770ab7
CY		 148 return true;
149
150 segno = GET_SEGNO(sbi, blkaddr);
151 offset = GET_BLKOFF_FROM_SEG0(sbi, blkaddr);
152 se = get_seg_entry(sbi, segno);
153
154 exist = f2fs_test_bit(offset, se->cur_valid_map);
0ef4ca04
CY		 155 if (exist && type == DATA_GENERIC_ENHANCE_UPDATE) {
156 f2fs_err(sbi, "Inconsistent error blkaddr:%u, sit bitmap:%d",
157 blkaddr, exist);
158 set_sbi_flag(sbi, SBI_NEED_FSCK);
159 return exist;
160 }
161
93770ab7 162 if (!exist && type == DATA_GENERIC_ENHANCE) {
dcbb4c10
JP		 163 f2fs_err(sbi, "Inconsistent error blkaddr:%u, sit bitmap:%d",
164 blkaddr, exist);
93770ab7 165 set_sbi_flag(sbi, SBI_NEED_FSCK);
dc2f78e2 166 dump_stack();
93770ab7
CY		 167 }
168 return exist;
169}
170
e1da7872 171bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi,
4d57b86d 172 block_t blkaddr, int type)
662befda 173{
18792e64
CY		 174 if (time_to_inject(sbi, FAULT_BLKADDR)) {
175 f2fs_show_injection_info(sbi, FAULT_BLKADDR);
176 return false;
177 }
178
662befda
CY		 179 switch (type) {
180 case META_NAT:
66b00c18 181 break;
662befda 182 case META_SIT:
66b00c18
CY		 183 if (unlikely(blkaddr >= SIT_BLK_CNT(sbi)))
184 return false;
185 break;
81c1a0f1 186 case META_SSA:
66b00c18
CY		 187 if (unlikely(blkaddr >= MAIN_BLKADDR(sbi) ||
188 blkaddr < SM_I(sbi)->ssa_blkaddr))
189 return false;
190 break;
662befda 191 case META_CP:
66b00c18
CY		 192 if (unlikely(blkaddr >= SIT_I(sbi)->sit_base_addr ||
193 blkaddr < __start_cp_addr(sbi)))
194 return false;
195 break;
4c521f49 196 case META_POR:
93770ab7
CY		 197 if (unlikely(blkaddr >= MAX_BLKADDR(sbi) ||
198 blkaddr < MAIN_BLKADDR(sbi)))
199 return false;
200 break;
e1da7872 201 case DATA_GENERIC:
93770ab7
CY		 202 case DATA_GENERIC_ENHANCE:
203 case DATA_GENERIC_ENHANCE_READ:
0ef4ca04 204 case DATA_GENERIC_ENHANCE_UPDATE:
66b00c18 205 if (unlikely(blkaddr >= MAX_BLKADDR(sbi) ||
93770ab7 206 blkaddr < MAIN_BLKADDR(sbi))) {
dcbb4c10
JP		 207 f2fs_warn(sbi, "access invalid blkaddr:%u",
208 blkaddr);
93770ab7 209 set_sbi_flag(sbi, SBI_NEED_FSCK);
dc2f78e2 210 dump_stack();
66b00c18 211 return false;
93770ab7
CY		 212 } else {
213 return __is_bitmap_valid(sbi, blkaddr, type);
c9b60788 214 }
66b00c18 215 break;
e1da7872
CY		 216 case META_GENERIC:
217 if (unlikely(blkaddr < SEG0_BLKADDR(sbi) ||
218 blkaddr >= MAIN_BLKADDR(sbi)))
219 return false;
220 break;
662befda
CY		 221 default:
222 BUG();
223 }
66b00c18
CY		 224
225 return true;
662befda
CY		 226}
227
228/*
7a88ddb5 229 * Readahead CP/NAT/SIT/SSA/POR pages
662befda 230 */
4d57b86d 231int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
26879fb1 232 int type, bool sync)
662befda 233{
662befda 234 struct page *page;
4c521f49 235 block_t blkno = start;
662befda 236 struct f2fs_io_info fio = {
05ca3632 237 .sbi = sbi,
662befda 238 .type = META,
04d328de 239 .op = REQ_OP_READ,
70fd7614 240 .op_flags = sync ? (REQ_META | REQ_PRIO) : REQ_RAHEAD,
4375a336 241 .encrypted_page = NULL,
fb830fc5 242 .in_list = false,
6dc3a126 243 .is_por = (type == META_POR),
662befda 244 };
e9f5b8b8 245 struct blk_plug plug;
ce4c638c 246 int err;
662befda 247
2b947003 248 if (unlikely(type == META_POR))
04d328de 249 fio.op_flags &= ~REQ_META;
2b947003 250
e9f5b8b8 251 blk_start_plug(&plug);
662befda 252 for (; nrpages-- > 0; blkno++) {
662befda 253
e1da7872 254 if (!f2fs_is_valid_blkaddr(sbi, blkno, type))
66b00c18
CY		 255 goto out;
256
662befda
CY		 257 switch (type) {
258 case META_NAT:
66b00c18
CY		 259 if (unlikely(blkno >=
260 NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid)))
662befda 261 blkno = 0;
66b00c18 262 /* get nat block addr */
7a9d7548 263 fio.new_blkaddr = current_nat_addr(sbi,
662befda
CY		 264 blkno * NAT_ENTRY_PER_BLOCK);
265 break;
266 case META_SIT:
6a257471
CY		 267 if (unlikely(blkno >= TOTAL_SEGS(sbi)))
268 goto out;
662befda 269 /* get sit block addr */
7a9d7548 270 fio.new_blkaddr = current_sit_addr(sbi,
662befda 271 blkno * SIT_ENTRY_PER_BLOCK);
662befda 272 break;
81c1a0f1 273 case META_SSA:
662befda 274 case META_CP:
4c521f49 275 case META_POR:
7a9d7548 276 fio.new_blkaddr = blkno;
662befda
CY		 277 break;
278 default:
279 BUG();
280 }
281
300e129c
JK		 282 page = f2fs_grab_cache_page(META_MAPPING(sbi),
283 fio.new_blkaddr, false);
662befda
CY		 284 if (!page)
285 continue;
286 if (PageUptodate(page)) {
662befda
CY		 287 f2fs_put_page(page, 1);
288 continue;
289 }
290
05ca3632 291 fio.page = page;
ce4c638c
CY		 292 err = f2fs_submit_page_bio(&fio);
293 f2fs_put_page(page, err ? 1 : 0);
8b83ac81
CY		 294
295 if (!err)
34a23525
CY		 296 f2fs_update_iostat(sbi, NULL, FS_META_READ_IO,
297 F2FS_BLKSIZE);
662befda
CY		 298 }
299out:
e9f5b8b8 300 blk_finish_plug(&plug);
662befda
CY		 301 return blkno - start;
302}
303
430f163b
CY		 304void f2fs_ra_meta_pages_cond(struct f2fs_sb_info *sbi, pgoff_t index,
305 unsigned int ra_blocks)
635aee1f
CY		 306{
307 struct page *page;
308 bool readahead = false;
309
430f163b
CY		 310 if (ra_blocks == RECOVERY_MIN_RA_BLOCKS)
311 return;
312
635aee1f 313 page = find_get_page(META_MAPPING(sbi), index);
4da7bf5a 314 if (!page || !PageUptodate(page))
635aee1f
CY		 315 readahead = true;
316 f2fs_put_page(page, 0);
317
318 if (readahead)
430f163b 319 f2fs_ra_meta_pages(sbi, index, ra_blocks, META_POR, true);
635aee1f
CY		 320}
321
b0af6d49
CY		 322static int __f2fs_write_meta_page(struct page *page,
323 struct writeback_control *wbc,
324 enum iostat_type io_type)
127e670a 325{
4081363f 326 struct f2fs_sb_info *sbi = F2FS_P_SB(page);
127e670a 327
ecda0de3
CY		 328 trace_f2fs_writepage(page, META);
329
af697c0f
JK		 330 if (unlikely(f2fs_cp_error(sbi)))
331 goto redirty_out;
caf0047e 332 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
cfb271d4 333 goto redirty_out;
857dc4e0 334 if (wbc->for_reclaim && page->index < GET_SUM_BLOCK(sbi, 0))
cfb271d4 335 goto redirty_out;
127e670a 336
4d57b86d 337 f2fs_do_write_meta_page(sbi, page, io_type);
577e3495 338 dec_page_count(sbi, F2FS_DIRTY_META);
0c3a5797
CY		 339
340 if (wbc->for_reclaim)
bab475c5 341 f2fs_submit_merged_write_cond(sbi, NULL, page, 0, META);
0c3a5797 342
577e3495 343 unlock_page(page);
857dc4e0 344
0c3a5797 345 if (unlikely(f2fs_cp_error(sbi)))
b9109b0e 346 f2fs_submit_merged_write(sbi, META);
0c3a5797 347
577e3495 348 return 0;
cfb271d4
CY		 349
350redirty_out:
76f60268 351 redirty_page_for_writepage(wbc, page);
cfb271d4 352 return AOP_WRITEPAGE_ACTIVATE;
127e670a
JK		 353}
354
b0af6d49
CY		 355static int f2fs_write_meta_page(struct page *page,
356 struct writeback_control *wbc)
357{
358 return __f2fs_write_meta_page(page, wbc, FS_META_IO);
359}
360
127e670a
JK		 361static int f2fs_write_meta_pages(struct address_space *mapping,
362 struct writeback_control *wbc)
363{
4081363f 364 struct f2fs_sb_info *sbi = F2FS_M_SB(mapping);
50c8cdb3 365 long diff, written;
127e670a 366
0771fcc7
CY		 367 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
368 goto skip_write;
369
5459aa97 370 /* collect a number of dirty meta pages and write together */
812a9597
JK		 371 if (wbc->sync_mode != WB_SYNC_ALL &&
372 get_pages(sbi, F2FS_DIRTY_META) <
373 nr_pages_to_skip(sbi, META))
d3baf95d 374 goto skip_write;
127e670a 375
a29d0e0b 376 /* if locked failed, cp will flush dirty pages instead */
e4544b63 377 if (!f2fs_down_write_trylock(&sbi->cp_global_sem))
a29d0e0b 378 goto skip_write;
d31c7c3f 379
a29d0e0b 380 trace_f2fs_writepages(mapping->host, wbc, META);
50c8cdb3 381 diff = nr_pages_to_write(sbi, META, wbc);
4d57b86d 382 written = f2fs_sync_meta_pages(sbi, META, wbc->nr_to_write, FS_META_IO);
e4544b63 383 f2fs_up_write(&sbi->cp_global_sem);
50c8cdb3 384 wbc->nr_to_write = max((long)0, wbc->nr_to_write - written - diff);
127e670a 385 return 0;
d3baf95d
JK		 386
387skip_write:
388 wbc->pages_skipped += get_pages(sbi, F2FS_DIRTY_META);
d31c7c3f 389 trace_f2fs_writepages(mapping->host, wbc, META);
d3baf95d 390 return 0;
127e670a
JK		 391}
392
4d57b86d 393long f2fs_sync_meta_pages(struct f2fs_sb_info *sbi, enum page_type type,
b0af6d49 394 long nr_to_write, enum iostat_type io_type)
127e670a 395{
9df27d98 396 struct address_space *mapping = META_MAPPING(sbi);
028a63a6 397 pgoff_t index = 0, prev = ULONG_MAX;
127e670a
JK		 398 struct pagevec pvec;
399 long nwritten = 0;
028a63a6 400 int nr_pages;
127e670a
JK		 401 struct writeback_control wbc = {
402 .for_reclaim = 0,
403 };
e9f5b8b8 404 struct blk_plug plug;
127e670a 405
86679820 406 pagevec_init(&pvec);
127e670a 407
e9f5b8b8
CY		 408 blk_start_plug(&plug);
409
028a63a6 410 while ((nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
67fd707f 411 PAGECACHE_TAG_DIRTY))) {
028a63a6 412 int i;
127e670a
JK		 413
414 for (i = 0; i < nr_pages; i++) {
415 struct page *page = pvec.pages[i];
203681f6 416
80dd9c0e 417 if (prev == ULONG_MAX)
6066d8cd
JK		 418 prev = page->index - 1;
419 if (nr_to_write != LONG_MAX && page->index != prev + 1) {
420 pagevec_release(&pvec);
421 goto stop;
422 }
423
127e670a 424 lock_page(page);
203681f6
JK		 425
426 if (unlikely(page->mapping != mapping)) {
427continue_unlock:
428 unlock_page(page);
429 continue;
430 }
431 if (!PageDirty(page)) {
432 /* someone wrote it for us */
433 goto continue_unlock;
434 }
435
bae0ee7a 436 f2fs_wait_on_page_writeback(page, META, true, true);
fa3d2bdf 437
203681f6
JK		 438 if (!clear_page_dirty_for_io(page))
439 goto continue_unlock;
440
b0af6d49 441 if (__f2fs_write_meta_page(page, &wbc, io_type)) {
577e3495
JK		 442 unlock_page(page);
443 break;
444 }
cfb271d4 445 nwritten++;
6066d8cd 446 prev = page->index;
cfb271d4 447 if (unlikely(nwritten >= nr_to_write))
127e670a
JK		 448 break;
449 }
450 pagevec_release(&pvec);
451 cond_resched();
452 }
6066d8cd 453stop:
127e670a 454 if (nwritten)
b9109b0e 455 f2fs_submit_merged_write(sbi, type);
127e670a 456
e9f5b8b8
CY		 457 blk_finish_plug(&plug);
458
127e670a
JK		 459 return nwritten;
460}
461
1d9ac659
MWO		 462static bool f2fs_dirty_meta_folio(struct address_space *mapping,
463 struct folio *folio)
127e670a 464{
1d9ac659
MWO		 465 trace_f2fs_set_page_dirty(&folio->page, META);
466
467 if (!folio_test_uptodate(folio))
468 folio_mark_uptodate(folio);
9b7eadd9 469 if (filemap_dirty_folio(mapping, folio)) {
29c87793 470 inc_page_count(F2FS_M_SB(mapping), F2FS_DIRTY_META);
1d9ac659
MWO		 471 set_page_private_reference(&folio->page);
472 return true;
127e670a 473 }
1d9ac659 474 return false;
127e670a
JK		 475}
476
477const struct address_space_operations f2fs_meta_aops = {
478 .writepage = f2fs_write_meta_page,
479 .writepages = f2fs_write_meta_pages,
1d9ac659 480 .dirty_folio = f2fs_dirty_meta_folio,
91503996 481 .invalidate_folio = f2fs_invalidate_folio,
c26cd045 482 .release_folio = f2fs_release_folio,
1d5b9bd6 483 .migrate_folio = filemap_migrate_folio,
127e670a
JK		 484};
485
39d787be
CY		 486static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino,
487 unsigned int devidx, int type)
953e6cc6 488{
67298804 489 struct inode_management *im = &sbi->im[type];
4b106518 490 struct ino_entry *e = NULL, *new = NULL;
80c54505 491
4b106518
CY		 492 if (type == FLUSH_INO) {
493 rcu_read_lock();
494 e = radix_tree_lookup(&im->ino_root, ino);
495 rcu_read_unlock();
496 }
497
498retry:
499 if (!e)
32410577
CY		 500 new = f2fs_kmem_cache_alloc(ino_entry_slab,
501 GFP_NOFS, true, NULL);
19526d74 502
80c54505 503 radix_tree_preload(GFP_NOFS | __GFP_NOFAIL);
769ec6e5 504
67298804 505 spin_lock(&im->ino_lock);
67298804 506 e = radix_tree_lookup(&im->ino_root, ino);
39efac41 507 if (!e) {
4b106518
CY		 508 if (!new) {
509 spin_unlock(&im->ino_lock);
510 goto retry;
511 }
512 e = new;
19526d74
CY		 513 if (unlikely(radix_tree_insert(&im->ino_root, ino, e)))
514 f2fs_bug_on(sbi, 1);
515
39efac41
JK		 516 memset(e, 0, sizeof(struct ino_entry));
517 e->ino = ino;
953e6cc6 518
67298804 519 list_add_tail(&e->list, &im->ino_list);
8c402946 520 if (type != ORPHAN_INO)
67298804 521 im->ino_num++;
39efac41 522 }
39d787be
CY		 523
524 if (type == FLUSH_INO)
525 f2fs_set_bit(devidx, (char *)&e->dirty_device);
526
67298804 527 spin_unlock(&im->ino_lock);
769ec6e5 528 radix_tree_preload_end();
80c54505 529
4b106518
CY		 530 if (new && e != new)
531 kmem_cache_free(ino_entry_slab, new);
953e6cc6
JK		 532}
533
6451e041 534static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
953e6cc6 535{
67298804 536 struct inode_management *im = &sbi->im[type];
6451e041 537 struct ino_entry *e;
953e6cc6 538
67298804
CY		 539 spin_lock(&im->ino_lock);
540 e = radix_tree_lookup(&im->ino_root, ino);
39efac41
JK		 541 if (e) {
542 list_del(&e->list);
67298804
CY		 543 radix_tree_delete(&im->ino_root, ino);
544 im->ino_num--;
545 spin_unlock(&im->ino_lock);
39efac41
JK		 546 kmem_cache_free(ino_entry_slab, e);
547 return;
953e6cc6 548 }
67298804 549 spin_unlock(&im->ino_lock);
953e6cc6
JK		 550}
551
4d57b86d 552void f2fs_add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
fff04f90
JK		 553{
554 /* add new dirty ino entry into list */
39d787be 555 __add_ino_entry(sbi, ino, 0, type);
fff04f90
JK		 556}
557
4d57b86d 558void f2fs_remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
fff04f90
JK		 559{
560 /* remove dirty ino entry from list */
561 __remove_ino_entry(sbi, ino, type);
562}
563
1f07cc58 564/* mode should be APPEND_INO, UPDATE_INO or TRANS_DIR_INO */
4d57b86d 565bool f2fs_exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
fff04f90 566{
67298804 567 struct inode_management *im = &sbi->im[mode];
fff04f90 568 struct ino_entry *e;
67298804
CY		 569
570 spin_lock(&im->ino_lock);
571 e = radix_tree_lookup(&im->ino_root, ino);
572 spin_unlock(&im->ino_lock);
fff04f90
JK		 573 return e ? true : false;
574}
575
4d57b86d 576void f2fs_release_ino_entry(struct f2fs_sb_info *sbi, bool all)
fff04f90
JK		 577{
578 struct ino_entry *e, *tmp;
579 int i;
580
39d787be 581 for (i = all ? ORPHAN_INO : APPEND_INO; i < MAX_INO_ENTRY; i++) {
67298804
CY		 582 struct inode_management *im = &sbi->im[i];
583
584 spin_lock(&im->ino_lock);
585 list_for_each_entry_safe(e, tmp, &im->ino_list, list) {
fff04f90 586 list_del(&e->list);
67298804 587 radix_tree_delete(&im->ino_root, e->ino);
fff04f90 588 kmem_cache_free(ino_entry_slab, e);
67298804 589 im->ino_num--;
fff04f90 590 }
67298804 591 spin_unlock(&im->ino_lock);
fff04f90
JK		 592 }
593}
594
4d57b86d 595void f2fs_set_dirty_device(struct f2fs_sb_info *sbi, nid_t ino,
39d787be
CY		 596 unsigned int devidx, int type)
597{
598 __add_ino_entry(sbi, ino, devidx, type);
599}
600
4d57b86d 601bool f2fs_is_dirty_device(struct f2fs_sb_info *sbi, nid_t ino,
39d787be
CY		 602 unsigned int devidx, int type)
603{
604 struct inode_management *im = &sbi->im[type];
605 struct ino_entry *e;
606 bool is_dirty = false;
607
608 spin_lock(&im->ino_lock);
609 e = radix_tree_lookup(&im->ino_root, ino);
610 if (e && f2fs_test_bit(devidx, (char *)&e->dirty_device))
611 is_dirty = true;
612 spin_unlock(&im->ino_lock);
613 return is_dirty;
614}
615
4d57b86d 616int f2fs_acquire_orphan_inode(struct f2fs_sb_info *sbi)
127e670a 617{
67298804 618 struct inode_management *im = &sbi->im[ORPHAN_INO];
127e670a
JK		 619 int err = 0;
620
67298804 621 spin_lock(&im->ino_lock);
cb78942b 622
1ecc0c5c 623 if (time_to_inject(sbi, FAULT_ORPHAN)) {
cb78942b 624 spin_unlock(&im->ino_lock);
c45d6002 625 f2fs_show_injection_info(sbi, FAULT_ORPHAN);
cb78942b
JK		 626 return -ENOSPC;
627 }
7fa750a1 628
67298804 629 if (unlikely(im->ino_num >= sbi->max_orphans))
127e670a 630 err = -ENOSPC;
cbd56e7d 631 else
67298804
CY		 632 im->ino_num++;
633 spin_unlock(&im->ino_lock);
0d47c1ad 634
127e670a
JK		 635 return err;
636}
637
4d57b86d 638void f2fs_release_orphan_inode(struct f2fs_sb_info *sbi)
cbd56e7d 639{
67298804
CY		 640 struct inode_management *im = &sbi->im[ORPHAN_INO];
641
642 spin_lock(&im->ino_lock);
643 f2fs_bug_on(sbi, im->ino_num == 0);
644 im->ino_num--;
645 spin_unlock(&im->ino_lock);
cbd56e7d
JK		 646}
647
4d57b86d 648void f2fs_add_orphan_inode(struct inode *inode)
127e670a 649{
39efac41 650 /* add new orphan ino entry into list */
39d787be 651 __add_ino_entry(F2FS_I_SB(inode), inode->i_ino, 0, ORPHAN_INO);
4d57b86d 652 f2fs_update_inode_page(inode);
127e670a
JK		 653}
654
4d57b86d 655void f2fs_remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
127e670a 656{
953e6cc6 657 /* remove orphan entry from orphan list */
6451e041 658 __remove_ino_entry(sbi, ino, ORPHAN_INO);
127e670a
JK		 659}
660
8c14bfad 661static int recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
127e670a 662{
8c14bfad 663 struct inode *inode;
5905f9af 664 struct node_info ni;
76a45e3c 665 int err;
8c14bfad 666
5905f9af 667 inode = f2fs_iget_retry(sbi->sb, ino);
8c14bfad
CY		 668 if (IS_ERR(inode)) {
669 /*
670 * there should be a bug that we can't find the entry
671 * to orphan inode.
672 */
673 f2fs_bug_on(sbi, PTR_ERR(inode) == -ENOENT);
674 return PTR_ERR(inode);
675 }
676
10a26878 677 err = f2fs_dquot_initialize(inode);
a515d12f
SY		 678 if (err) {
679 iput(inode);
0f9ec2a8 680 goto err_out;
a515d12f 681 }
0f9ec2a8 682
127e670a
JK		 683 clear_nlink(inode);
684
685 /* truncate all the data during iput */
686 iput(inode);
5905f9af 687
a9419b63 688 err = f2fs_get_node_info(sbi, ino, &ni, false);
7735730d
CY		 689 if (err)
690 goto err_out;
5905f9af
JK		 691
692 /* ENOMEM was fully retried in f2fs_evict_inode. */
693 if (ni.blk_addr != NULL_ADDR) {
0f9ec2a8
JK		 694 err = -EIO;
695 goto err_out;
5905f9af 696 }
8c14bfad 697 return 0;
0f9ec2a8
JK		 698
699err_out:
700 set_sbi_flag(sbi, SBI_NEED_FSCK);
dcbb4c10
JP		 701 f2fs_warn(sbi, "%s: orphan failed (ino=%x), run fsck to fix.",
702 __func__, ino);
0f9ec2a8 703 return err;
127e670a
JK		 704}
705
4d57b86d 706int f2fs_recover_orphan_inodes(struct f2fs_sb_info *sbi)
127e670a 707{
3c642985 708 block_t start_blk, orphan_blocks, i, j;
4b2414d0
CY		 709 unsigned int s_flags = sbi->sb->s_flags;
710 int err = 0;
ea676733
JK		 711#ifdef CONFIG_QUOTA
712 int quota_enabled;
713#endif
127e670a 714
aaec2b1d 715 if (!is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG))
8c14bfad 716 return 0;
127e670a 717
b61af314 718 if (bdev_read_only(sbi->sb->s_bdev)) {
dcbb4c10 719 f2fs_info(sbi, "write access unavailable, skipping orphan cleanup");
b61af314
CY		 720 return 0;
721 }
722
1751e8a6 723 if (s_flags & SB_RDONLY) {
dcbb4c10 724 f2fs_info(sbi, "orphan cleanup on readonly fs");
1751e8a6 725 sbi->sb->s_flags &= ~SB_RDONLY;
4b2414d0
CY		 726 }
727
728#ifdef CONFIG_QUOTA
76cf05d7
SY		 729 /*
730 * Turn on quotas which were not enabled for read-only mounts if
731 * filesystem has quota feature, so that they are updated correctly.
732 */
1751e8a6 733 quota_enabled = f2fs_enable_quota_files(sbi, s_flags & SB_RDONLY);
4b2414d0
CY		 734#endif
735
55141486 736 start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi);
3c642985 737 orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi);
127e670a 738
4d57b86d 739 f2fs_ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP, true);
662befda 740
3c642985 741 for (i = 0; i < orphan_blocks; i++) {
7735730d 742 struct page *page;
127e670a
JK		 743 struct f2fs_orphan_block *orphan_blk;
744
7735730d
CY		 745 page = f2fs_get_meta_page(sbi, start_blk + i);
746 if (IS_ERR(page)) {
747 err = PTR_ERR(page);
748 goto out;
749 }
750
127e670a
JK		 751 orphan_blk = (struct f2fs_orphan_block *)page_address(page);
752 for (j = 0; j < le32_to_cpu(orphan_blk->entry_count); j++) {
753 nid_t ino = le32_to_cpu(orphan_blk->ino[j]);
5f029c04 754
8c14bfad
CY		 755 err = recover_orphan_inode(sbi, ino);
756 if (err) {
757 f2fs_put_page(page, 1);
4b2414d0 758 goto out;
8c14bfad 759 }
127e670a
JK		 760 }
761 f2fs_put_page(page, 1);
762 }
763 /* clear Orphan Flag */
aaec2b1d 764 clear_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG);
4b2414d0 765out:
1378752b
CY		 766 set_sbi_flag(sbi, SBI_IS_RECOVERED);
767
4b2414d0
CY		 768#ifdef CONFIG_QUOTA
769 /* Turn quotas off */
ea676733
JK		 770 if (quota_enabled)
771 f2fs_quota_off_umount(sbi->sb);
4b2414d0 772#endif
1751e8a6 773 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */
4b2414d0
CY		 774
775 return err;
127e670a
JK		 776}
777
778static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
779{
502c6e0b 780 struct list_head *head;
127e670a 781 struct f2fs_orphan_block *orphan_blk = NULL;
127e670a 782 unsigned int nentries = 0;
bd936f84 783 unsigned short index = 1;
8c402946 784 unsigned short orphan_blocks;
4531929e 785 struct page *page = NULL;
6451e041 786 struct ino_entry *orphan = NULL;
67298804 787 struct inode_management *im = &sbi->im[ORPHAN_INO];
127e670a 788
67298804 789 orphan_blocks = GET_ORPHAN_BLOCKS(im->ino_num);
8c402946 790
d6c67a4f
JK		 791 /*
792 * we don't need to do spin_lock(&im->ino_lock) here, since all the
793 * orphan inode operations are covered under f2fs_lock_op().
794 * And, spin_lock should be avoided due to page operations below.
795 */
67298804 796 head = &im->ino_list;
127e670a
JK		 797
798 /* loop for each orphan inode entry and write them in Jornal block */
502c6e0b
GZ		 799 list_for_each_entry(orphan, head, list) {
800 if (!page) {
4d57b86d 801 page = f2fs_grab_meta_page(sbi, start_blk++);
502c6e0b
GZ		 802 orphan_blk =
803 (struct f2fs_orphan_block *)page_address(page);
804 memset(orphan_blk, 0, sizeof(*orphan_blk));
805 }
127e670a 806
36795567 807 orphan_blk->ino[nentries++] = cpu_to_le32(orphan->ino);
127e670a 808
36795567 809 if (nentries == F2FS_ORPHANS_PER_BLOCK) {
127e670a
JK		 810 /*
811 * an orphan block is full of 1020 entries,
812 * then we need to flush current orphan blocks
813 * and bring another one in memory
814 */
815 orphan_blk->blk_addr = cpu_to_le16(index);
816 orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
817 orphan_blk->entry_count = cpu_to_le32(nentries);
818 set_page_dirty(page);
819 f2fs_put_page(page, 1);
820 index++;
127e670a
JK		 821 nentries = 0;
822 page = NULL;
823 }
502c6e0b 824 }
127e670a 825
502c6e0b
GZ		 826 if (page) {
827 orphan_blk->blk_addr = cpu_to_le16(index);
828 orphan_blk->blk_count = cpu_to_le16(orphan_blocks);
829 orphan_blk->entry_count = cpu_to_le32(nentries);
830 set_page_dirty(page);
831 f2fs_put_page(page, 1);
127e670a 832 }
127e670a
JK		 833}
834
d7eb8f1c
CY		 835static __u32 f2fs_checkpoint_chksum(struct f2fs_sb_info *sbi,
836 struct f2fs_checkpoint *ckpt)
837{
838 unsigned int chksum_ofs = le32_to_cpu(ckpt->checksum_offset);
839 __u32 chksum;
840
841 chksum = f2fs_crc32(sbi, ckpt, chksum_ofs);
842 if (chksum_ofs < CP_CHKSUM_OFFSET) {
843 chksum_ofs += sizeof(chksum);
844 chksum = f2fs_chksum(sbi, chksum, (__u8 *)ckpt + chksum_ofs,
845 F2FS_BLKSIZE - chksum_ofs);
846 }
847 return chksum;
848}
849
fc0065ad
TY		 850static int get_checkpoint_version(struct f2fs_sb_info *sbi, block_t cp_addr,
851 struct f2fs_checkpoint **cp_block, struct page **cp_page,
852 unsigned long long *version)
127e670a 853{
fc0065ad 854 size_t crc_offset = 0;
d7eb8f1c 855 __u32 crc;
127e670a 856
4d57b86d 857 *cp_page = f2fs_get_meta_page(sbi, cp_addr);
7735730d
CY		 858 if (IS_ERR(*cp_page))
859 return PTR_ERR(*cp_page);
860
fc0065ad 861 *cp_block = (struct f2fs_checkpoint *)page_address(*cp_page);
127e670a 862
fc0065ad 863 crc_offset = le32_to_cpu((*cp_block)->checksum_offset);
d7eb8f1c
CY		 864 if (crc_offset < CP_MIN_CHKSUM_OFFSET ||
865 crc_offset > CP_CHKSUM_OFFSET) {
d3f07c04 866 f2fs_put_page(*cp_page, 1);
dcbb4c10 867 f2fs_warn(sbi, "invalid crc_offset: %zu", crc_offset);
fc0065ad
TY		 868 return -EINVAL;
869 }
127e670a 870
d7eb8f1c
CY		 871 crc = f2fs_checkpoint_chksum(sbi, *cp_block);
872 if (crc != cur_cp_crc(*cp_block)) {
d3f07c04 873 f2fs_put_page(*cp_page, 1);
dcbb4c10 874 f2fs_warn(sbi, "invalid crc value");
fc0065ad
TY		 875 return -EINVAL;
876 }
127e670a 877
fc0065ad
TY		 878 *version = cur_cp_version(*cp_block);
879 return 0;
880}
127e670a 881
fc0065ad
TY		 882static struct page *validate_checkpoint(struct f2fs_sb_info *sbi,
883 block_t cp_addr, unsigned long long *version)
884{
885 struct page *cp_page_1 = NULL, *cp_page_2 = NULL;
886 struct f2fs_checkpoint *cp_block = NULL;
887 unsigned long long cur_version = 0, pre_version = 0;
5b5b4f85 888 unsigned int cp_blocks;
fc0065ad 889 int err;
127e670a 890
fc0065ad
TY		 891 err = get_checkpoint_version(sbi, cp_addr, &cp_block,
892 &cp_page_1, version);
893 if (err)
d3f07c04 894 return NULL;
c9b60788 895
5b5b4f85
CY		 896 cp_blocks = le32_to_cpu(cp_block->cp_pack_total_block_count);
897
898 if (cp_blocks > sbi->blocks_per_seg || cp_blocks <= F2FS_CP_PACKS) {
dcbb4c10
JP		 899 f2fs_warn(sbi, "invalid cp_pack_total_block_count:%u",
900 le32_to_cpu(cp_block->cp_pack_total_block_count));
d3f07c04 901 goto invalid_cp;
c9b60788 902 }
fc0065ad 903 pre_version = *version;
127e670a 904
5b5b4f85 905 cp_addr += cp_blocks - 1;
fc0065ad
TY		 906 err = get_checkpoint_version(sbi, cp_addr, &cp_block,
907 &cp_page_2, version);
908 if (err)
d3f07c04 909 goto invalid_cp;
fc0065ad 910 cur_version = *version;
127e670a
JK		 911
912 if (cur_version == pre_version) {
913 *version = cur_version;
914 f2fs_put_page(cp_page_2, 1);
915 return cp_page_1;
916 }
127e670a 917 f2fs_put_page(cp_page_2, 1);
d3f07c04 918invalid_cp:
127e670a
JK		 919 f2fs_put_page(cp_page_1, 1);
920 return NULL;
921}
922
4d57b86d 923int f2fs_get_valid_checkpoint(struct f2fs_sb_info *sbi)
127e670a
JK		 924{
925 struct f2fs_checkpoint *cp_block;
926 struct f2fs_super_block *fsb = sbi->raw_super;
927 struct page *cp1, *cp2, *cur_page;
928 unsigned long blk_size = sbi->blocksize;
929 unsigned long long cp1_version = 0, cp2_version = 0;
930 unsigned long long cp_start_blk_no;
55141486 931 unsigned int cp_blks = 1 + __cp_payload(sbi);
1dbe4152
CL		 932 block_t cp_blk_no;
933 int i;
10f966bb 934 int err;
127e670a 935
0b6d4ca0
EB		 936 sbi->ckpt = f2fs_kvzalloc(sbi, array_size(blk_size, cp_blks),
937 GFP_KERNEL);
127e670a
JK		 938 if (!sbi->ckpt)
939 return -ENOMEM;
940 /*
941 * Finding out valid cp block involves read both
7a88ddb5 942 * sets( cp pack 1 and cp pack 2)
127e670a
JK		 943 */
944 cp_start_blk_no = le32_to_cpu(fsb->cp_blkaddr);
945 cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version);
946
947 /* The second checkpoint pack should start at the next segment */
f9a4e6df
JK		 948 cp_start_blk_no += ((unsigned long long)1) <<
949 le32_to_cpu(fsb->log_blocks_per_seg);
127e670a
JK		 950 cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version);
951
952 if (cp1 && cp2) {
953 if (ver_after(cp2_version, cp1_version))
954 cur_page = cp2;
955 else
956 cur_page = cp1;
957 } else if (cp1) {
958 cur_page = cp1;
959 } else if (cp2) {
960 cur_page = cp2;
961 } else {
10f966bb 962 err = -EFSCORRUPTED;
127e670a
JK		 963 goto fail_no_cp;
964 }
965
966 cp_block = (struct f2fs_checkpoint *)page_address(cur_page);
967 memcpy(sbi->ckpt, cp_block, blk_size);
968
8508e44a
JK		 969 if (cur_page == cp1)
970 sbi->cur_cp_pack = 1;
971 else
972 sbi->cur_cp_pack = 2;
984ec63c 973
e494c2f9 974 /* Sanity checking of checkpoint */
10f966bb
CY		 975 if (f2fs_sanity_check_ckpt(sbi)) {
976 err = -EFSCORRUPTED;
e494c2f9 977 goto free_fail_no_cp;
10f966bb 978 }
e494c2f9 979
1dbe4152
CL		 980 if (cp_blks <= 1)
981 goto done;
982
983 cp_blk_no = le32_to_cpu(fsb->cp_blkaddr);
984 if (cur_page == cp2)
985 cp_blk_no += 1 << le32_to_cpu(fsb->log_blocks_per_seg);
986
987 for (i = 1; i < cp_blks; i++) {
988 void *sit_bitmap_ptr;
989 unsigned char *ckpt = (unsigned char *)sbi->ckpt;
990
4d57b86d 991 cur_page = f2fs_get_meta_page(sbi, cp_blk_no + i);
10f966bb
CY		 992 if (IS_ERR(cur_page)) {
993 err = PTR_ERR(cur_page);
7735730d 994 goto free_fail_no_cp;
10f966bb 995 }
1dbe4152
CL		 996 sit_bitmap_ptr = page_address(cur_page);
997 memcpy(ckpt + i * blk_size, sit_bitmap_ptr, blk_size);
998 f2fs_put_page(cur_page, 1);
999 }
1000done:
127e670a
JK		 1001 f2fs_put_page(cp1, 1);
1002 f2fs_put_page(cp2, 1);
1003 return 0;
1004
a2125ff7
JK		 1005free_fail_no_cp:
1006 f2fs_put_page(cp1, 1);
1007 f2fs_put_page(cp2, 1);
127e670a 1008fail_no_cp:
5222595d 1009 kvfree(sbi->ckpt);
10f966bb 1010 return err;
127e670a
JK		 1011}
1012
c227f912 1013static void __add_dirty_inode(struct inode *inode, enum inode_type type)
127e670a 1014{
4081363f 1015 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
c227f912 1016 int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE;
127e670a 1017
91942321 1018 if (is_inode_flag_set(inode, flag))
2710fd7e 1019 return;
2d7b822a 1020
91942321 1021 set_inode_flag(inode, flag);
7bc155fe 1022 list_add_tail(&F2FS_I(inode)->dirty_list, &sbi->inode_list[type]);
33fbd510 1023 stat_inc_dirty_inode(sbi, type);
5deb8267
JK		 1024}
1025
c227f912 1026static void __remove_dirty_inode(struct inode *inode, enum inode_type type)
6ad7609a 1027{
c227f912 1028 int flag = (type == DIR_INODE) ? FI_DIRTY_DIR : FI_DIRTY_FILE;
6ad7609a 1029
91942321 1030 if (get_dirty_pages(inode) || !is_inode_flag_set(inode, flag))
6ad7609a
CY		 1031 return;
1032
91942321
JK		 1033 list_del_init(&F2FS_I(inode)->dirty_list);
1034 clear_inode_flag(inode, flag);
33fbd510 1035 stat_dec_dirty_inode(F2FS_I_SB(inode), type);
6ad7609a
CY		 1036}
1037
4f5e34f7 1038void f2fs_update_dirty_folio(struct inode *inode, struct folio *folio)
5deb8267 1039{
4081363f 1040 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
c227f912 1041 enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE;
5deb8267 1042
5ac9f36f
CY		 1043 if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) &&
1044 !S_ISLNK(inode->i_mode))
127e670a 1045 return;
7bd59381 1046
1c4bf763
JK		 1047 spin_lock(&sbi->inode_lock[type]);
1048 if (type != FILE_INODE || test_opt(sbi, DATA_FLUSH))
10aa97c3 1049 __add_dirty_inode(inode, type);
b951a4ec 1050 inode_inc_dirty_pages(inode);
1c4bf763
JK		 1051 spin_unlock(&sbi->inode_lock[type]);
1052
4f5e34f7 1053 set_page_private_reference(&folio->page);
5deb8267
JK		 1054}
1055
4d57b86d 1056void f2fs_remove_dirty_inode(struct inode *inode)
127e670a 1057{
4081363f 1058 struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
c227f912 1059 enum inode_type type = S_ISDIR(inode->i_mode) ? DIR_INODE : FILE_INODE;
127e670a 1060
c227f912
CY		 1061 if (!S_ISDIR(inode->i_mode) && !S_ISREG(inode->i_mode) &&
1062 !S_ISLNK(inode->i_mode))
127e670a
JK		 1063 return;
1064
10aa97c3
JK		 1065 if (type == FILE_INODE && !test_opt(sbi, DATA_FLUSH))
1066 return;
1067
c227f912
CY		 1068 spin_lock(&sbi->inode_lock[type]);
1069 __remove_dirty_inode(inode, type);
1070 spin_unlock(&sbi->inode_lock[type]);
74d0b917
JK		 1071}
1072
d80afefb
CY		 1073int f2fs_sync_dirty_inodes(struct f2fs_sb_info *sbi, enum inode_type type,
1074 bool from_cp)
127e670a 1075{
ce3b7d80 1076 struct list_head *head;
127e670a 1077 struct inode *inode;
2710fd7e 1078 struct f2fs_inode_info *fi;
4cf18537 1079 bool is_dir = (type == DIR_INODE);
4db08d01 1080 unsigned long ino = 0;
4cf18537
CY		 1081
1082 trace_f2fs_sync_dirty_inodes_enter(sbi->sb, is_dir,
1083 get_pages(sbi, is_dir ?
1084 F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA));
127e670a 1085retry:
9b664822
ZQ		 1086 if (unlikely(f2fs_cp_error(sbi))) {
1087 trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir,
1088 get_pages(sbi, is_dir ?
1089 F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA));
6d5a1495 1090 return -EIO;
9b664822 1091 }
af41d3ee 1092
c227f912 1093 spin_lock(&sbi->inode_lock[type]);
ce3b7d80 1094
c227f912 1095 head = &sbi->inode_list[type];
127e670a 1096 if (list_empty(head)) {
c227f912 1097 spin_unlock(&sbi->inode_lock[type]);
4cf18537
CY		 1098 trace_f2fs_sync_dirty_inodes_exit(sbi->sb, is_dir,
1099 get_pages(sbi, is_dir ?
1100 F2FS_DIRTY_DENTS : F2FS_DIRTY_DATA));
6d5a1495 1101 return 0;
127e670a 1102 }
939afa94 1103 fi = list_first_entry(head, struct f2fs_inode_info, dirty_list);
2710fd7e 1104 inode = igrab(&fi->vfs_inode);
c227f912 1105 spin_unlock(&sbi->inode_lock[type]);
127e670a 1106 if (inode) {
4db08d01
JK		 1107 unsigned long cur_ino = inode->i_ino;
1108
d80afefb
CY		 1109 if (from_cp)
1110 F2FS_I(inode)->cp_task = current;
1111 F2FS_I(inode)->wb_task = current;
b0af6d49 1112
87d6f890 1113 filemap_fdatawrite(inode->i_mapping);
b0af6d49 1114
d80afefb
CY		 1115 F2FS_I(inode)->wb_task = NULL;
1116 if (from_cp)
1117 F2FS_I(inode)->cp_task = NULL;
b0af6d49 1118
127e670a 1119 iput(inode);
4db08d01 1120 /* We need to give cpu to another writers. */
2a63531a 1121 if (ino == cur_ino)
4db08d01 1122 cond_resched();
2a63531a 1123 else
4db08d01 1124 ino = cur_ino;
127e670a
JK		 1125 } else {
1126 /*
1127 * We should submit bio, since it exists several
1128 * wribacking dentry pages in the freeing inode.
1129 */
b9109b0e 1130 f2fs_submit_merged_write(sbi, DATA);
7ecebe5e 1131 cond_resched();
127e670a
JK		 1132 }
1133 goto retry;
1134}
1135
0f18b462
JK		 1136int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi)
1137{
1138 struct list_head *head = &sbi->inode_list[DIRTY_META];
1139 struct inode *inode;
1140 struct f2fs_inode_info *fi;
1141 s64 total = get_pages(sbi, F2FS_DIRTY_IMETA);
1142
1143 while (total--) {
1144 if (unlikely(f2fs_cp_error(sbi)))
1145 return -EIO;
1146
1147 spin_lock(&sbi->inode_lock[DIRTY_META]);
1148 if (list_empty(head)) {
1149 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1150 return 0;
1151 }
939afa94 1152 fi = list_first_entry(head, struct f2fs_inode_info,
0f18b462
JK		 1153 gdirty_list);
1154 inode = igrab(&fi->vfs_inode);
1155 spin_unlock(&sbi->inode_lock[DIRTY_META]);
1156 if (inode) {
18340edc
JK		 1157 sync_inode_metadata(inode, 0);
1158
1159 /* it's on eviction */
1160 if (is_inode_flag_set(inode, FI_DIRTY_INODE))
4d57b86d 1161 f2fs_update_inode_page(inode);
0f18b462
JK		 1162 iput(inode);
1163 }
dee668c1 1164 }
0f18b462
JK		 1165 return 0;
1166}
1167
59c9081b
YH		 1168static void __prepare_cp_block(struct f2fs_sb_info *sbi)
1169{
1170 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1171 struct f2fs_nm_info *nm_i = NM_I(sbi);
1172 nid_t last_nid = nm_i->next_scan_nid;
1173
1174 next_free_nid(sbi, &last_nid);
1175 ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi));
1176 ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi));
1177 ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi));
1178 ckpt->next_free_nid = cpu_to_le32(last_nid);
1179}
1180
af033b2a
CY		 1181static bool __need_flush_quota(struct f2fs_sb_info *sbi)
1182{
db6ec53b
JK		 1183 bool ret = false;
1184
af033b2a
CY		 1185 if (!is_journalled_quota(sbi))
1186 return false;
db6ec53b 1187
e4544b63 1188 if (!f2fs_down_write_trylock(&sbi->quota_sem))
a5c00422 1189 return true;
db6ec53b
JK		 1190 if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH)) {
1191 ret = false;
1192 } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR)) {
1193 ret = false;
1194 } else if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_FLUSH)) {
1195 clear_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
1196 ret = true;
1197 } else if (get_pages(sbi, F2FS_DIRTY_QDATA)) {
1198 ret = true;
1199 }
e4544b63 1200 f2fs_up_write(&sbi->quota_sem);
db6ec53b 1201 return ret;
af033b2a
CY		 1202}
1203
0a8165d7 1204/*
127e670a
JK		 1205 * Freeze all the FS-operations for checkpoint.
1206 */
cf779cab 1207static int block_operations(struct f2fs_sb_info *sbi)
127e670a 1208{
127e670a
JK		 1209 struct writeback_control wbc = {
1210 .sync_mode = WB_SYNC_ALL,
1211 .nr_to_write = LONG_MAX,
1212 .for_reclaim = 0,
1213 };
af033b2a 1214 int err = 0, cnt = 0;
c718379b 1215
34c061ad
SL		 1216 /*
1217 * Let's flush inline_data in dirty node pages.
1218 */
1219 f2fs_flush_inline_data(sbi);
1220
af033b2a 1221retry_flush_quotas:
db6ec53b 1222 f2fs_lock_all(sbi);
af033b2a
CY		 1223 if (__need_flush_quota(sbi)) {
1224 int locked;
1225
1226 if (++cnt > DEFAULT_RETRY_QUOTA_FLUSH_COUNT) {
1227 set_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH);
db6ec53b 1228 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
af033b2a
CY		 1229 goto retry_flush_dents;
1230 }
db6ec53b 1231 f2fs_unlock_all(sbi);
af033b2a
CY		 1232
1233 /* only failed during mount/umount/freeze/quotactl */
1234 locked = down_read_trylock(&sbi->sb->s_umount);
1235 f2fs_quota_sync(sbi->sb, -1);
1236 if (locked)
1237 up_read(&sbi->sb->s_umount);
af033b2a
CY		 1238 cond_resched();
1239 goto retry_flush_quotas;
1240 }
1241
1242retry_flush_dents:
127e670a 1243 /* write all the dirty dentry pages */
127e670a 1244 if (get_pages(sbi, F2FS_DIRTY_DENTS)) {
e479556b 1245 f2fs_unlock_all(sbi);
d80afefb 1246 err = f2fs_sync_dirty_inodes(sbi, DIR_INODE, true);
6d5a1495 1247 if (err)
1f5f11a3 1248 return err;
30973883 1249 cond_resched();
af033b2a 1250 goto retry_flush_quotas;
127e670a
JK		 1251 }
1252
59c9081b
YH		 1253 /*
1254 * POR: we should ensure that there are no dirty node pages
1255 * until finishing nat/sit flush. inode->i_blocks can be updated.
1256 */
e4544b63 1257 f2fs_down_write(&sbi->node_change);
59c9081b 1258
0f18b462 1259 if (get_pages(sbi, F2FS_DIRTY_IMETA)) {
e4544b63 1260 f2fs_up_write(&sbi->node_change);
0f18b462
JK		 1261 f2fs_unlock_all(sbi);
1262 err = f2fs_sync_inode_meta(sbi);
1263 if (err)
1f5f11a3 1264 return err;
30973883 1265 cond_resched();
af033b2a 1266 goto retry_flush_quotas;
0f18b462
JK		 1267 }
1268
39936837 1269retry_flush_nodes:
e4544b63 1270 f2fs_down_write(&sbi->node_write);
127e670a
JK		 1271
1272 if (get_pages(sbi, F2FS_DIRTY_NODES)) {
e4544b63 1273 f2fs_up_write(&sbi->node_write);
c29fd0c0 1274 atomic_inc(&sbi->wb_sync_req[NODE]);
4d57b86d 1275 err = f2fs_sync_node_pages(sbi, &wbc, false, FS_CP_NODE_IO);
c29fd0c0 1276 atomic_dec(&sbi->wb_sync_req[NODE]);
6d5a1495 1277 if (err) {
e4544b63 1278 f2fs_up_write(&sbi->node_change);
cf779cab 1279 f2fs_unlock_all(sbi);
1f5f11a3 1280 return err;
cf779cab 1281 }
30973883 1282 cond_resched();
39936837 1283 goto retry_flush_nodes;
127e670a 1284 }
59c9081b
YH		 1285
1286 /*
1287 * sbi->node_change is used only for AIO write_begin path which produces
1288 * dirty node blocks and some checkpoint values by block allocation.
1289 */
1290 __prepare_cp_block(sbi);
e4544b63 1291 f2fs_up_write(&sbi->node_change);
cf779cab 1292 return err;
127e670a
JK		 1293}
1294
1295static void unblock_operations(struct f2fs_sb_info *sbi)
1296{
e4544b63 1297 f2fs_up_write(&sbi->node_write);
e479556b 1298 f2fs_unlock_all(sbi);
127e670a
JK		 1299}
1300
bf22c3cc 1301void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type)
fb51b5ef
CL		 1302{
1303 DEFINE_WAIT(wait);
1304
1305 for (;;) {
bf22c3cc 1306 if (!get_pages(sbi, type))
fb51b5ef
CL		 1307 break;
1308
af697c0f
JK		 1309 if (unlikely(f2fs_cp_error(sbi)))
1310 break;
1311
9c30df7c
JK		 1312 if (type == F2FS_DIRTY_META)
1313 f2fs_sync_meta_pages(sbi, META, LONG_MAX,
1314 FS_CP_META_IO);
1fd28018
JK		 1315 else if (type == F2FS_WB_CP_DATA)
1316 f2fs_submit_merged_write(sbi, DATA);
828add77
JK		 1317
1318 prepare_to_wait(&sbi->cp_wait, &wait, TASK_UNINTERRUPTIBLE);
5df7731f 1319 io_schedule_timeout(DEFAULT_IO_TIMEOUT);
fb51b5ef
CL		 1320 }
1321 finish_wait(&sbi->cp_wait, &wait);
1322}
1323
e4c5d848
JK		 1324static void update_ckpt_flags(struct f2fs_sb_info *sbi, struct cp_control *cpc)
1325{
1326 unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num;
1327 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
d1aa2453 1328 unsigned long flags;
e4c5d848 1329
94c821fb 1330 if (cpc->reason & CP_UMOUNT) {
b702c83e
CY		 1331 if (le32_to_cpu(ckpt->cp_pack_total_block_count) +
1332 NM_I(sbi)->nat_bits_blocks > sbi->blocks_per_seg) {
94c821fb
CY		 1333 clear_ckpt_flags(sbi, CP_NAT_BITS_FLAG);
1334 f2fs_notice(sbi, "Disable nat_bits due to no space");
1335 } else if (!is_set_ckpt_flags(sbi, CP_NAT_BITS_FLAG) &&
1336 f2fs_nat_bitmap_enabled(sbi)) {
1337 f2fs_enable_nat_bits(sbi);
1338 set_ckpt_flags(sbi, CP_NAT_BITS_FLAG);
1339 f2fs_notice(sbi, "Rebuild and enable nat_bits");
1340 }
1341 }
e4c5d848 1342
94c821fb 1343 spin_lock_irqsave(&sbi->cp_lock, flags);
22ad0b6a 1344
1f43e2ad
CY		 1345 if (cpc->reason & CP_TRIMMED)
1346 __set_ckpt_flags(ckpt, CP_TRIMMED_FLAG);
cd36d7a1
CY		 1347 else
1348 __clear_ckpt_flags(ckpt, CP_TRIMMED_FLAG);
1f43e2ad 1349
c473f1a9 1350 if (cpc->reason & CP_UMOUNT)
e4c5d848
JK		 1351 __set_ckpt_flags(ckpt, CP_UMOUNT_FLAG);
1352 else
1353 __clear_ckpt_flags(ckpt, CP_UMOUNT_FLAG);
1354
c473f1a9 1355 if (cpc->reason & CP_FASTBOOT)
e4c5d848
JK		 1356 __set_ckpt_flags(ckpt, CP_FASTBOOT_FLAG);
1357 else
1358 __clear_ckpt_flags(ckpt, CP_FASTBOOT_FLAG);
1359
1360 if (orphan_num)
1361 __set_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG);
1362 else
1363 __clear_ckpt_flags(ckpt, CP_ORPHAN_PRESENT_FLAG);
1364
c84ef3c5 1365 if (is_sbi_flag_set(sbi, SBI_NEED_FSCK))
e4c5d848
JK		 1366 __set_ckpt_flags(ckpt, CP_FSCK_FLAG);
1367
c84ef3c5
ST		 1368 if (is_sbi_flag_set(sbi, SBI_IS_RESIZEFS))
1369 __set_ckpt_flags(ckpt, CP_RESIZEFS_FLAG);
1370 else
1371 __clear_ckpt_flags(ckpt, CP_RESIZEFS_FLAG);
1372
4354994f
DR		 1373 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED))
1374 __set_ckpt_flags(ckpt, CP_DISABLED_FLAG);
1375 else
1376 __clear_ckpt_flags(ckpt, CP_DISABLED_FLAG);
1377
db610a64
JK		 1378 if (is_sbi_flag_set(sbi, SBI_CP_DISABLED_QUICK))
1379 __set_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG);
1380 else
1381 __clear_ckpt_flags(ckpt, CP_DISABLED_QUICK_FLAG);
1382
af033b2a
CY		 1383 if (is_sbi_flag_set(sbi, SBI_QUOTA_SKIP_FLUSH))
1384 __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG);
bc88ac96
JK		 1385 else
1386 __clear_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG);
af033b2a
CY		 1387
1388 if (is_sbi_flag_set(sbi, SBI_QUOTA_NEED_REPAIR))
1389 __set_ckpt_flags(ckpt, CP_QUOTA_NEED_FSCK_FLAG);
1390
e4c5d848
JK		 1391 /* set this flag to activate crc|cp_ver for recovery */
1392 __set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG);
f2367923 1393 __clear_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG);
e4c5d848 1394
d1aa2453 1395 spin_unlock_irqrestore(&sbi->cp_lock, flags);
e4c5d848
JK		 1396}
1397
46706d59
GX		 1398static void commit_checkpoint(struct f2fs_sb_info *sbi,
1399 void *src, block_t blk_addr)
1400{
1401 struct writeback_control wbc = {
1402 .for_reclaim = 0,
1403 };
1404
1405 /*
1406 * pagevec_lookup_tag and lock_page again will take
4d57b86d
CY		 1407 * some extra time. Therefore, f2fs_update_meta_pages and
1408 * f2fs_sync_meta_pages are combined in this function.
46706d59 1409 */
4d57b86d 1410 struct page *page = f2fs_grab_meta_page(sbi, blk_addr);
46706d59
GX		 1411 int err;
1412
bae0ee7a 1413 f2fs_wait_on_page_writeback(page, META, true, true);
8d64d365
CY		 1414
1415 memcpy(page_address(page), src, PAGE_SIZE);
1416
1417 set_page_dirty(page);
46706d59
GX		 1418 if (unlikely(!clear_page_dirty_for_io(page)))
1419 f2fs_bug_on(sbi, 1);
1420
1421 /* writeout cp pack 2 page */
1422 err = __f2fs_write_meta_page(page, &wbc, FS_CP_META_IO);
af697c0f
JK		 1423 if (unlikely(err && f2fs_cp_error(sbi))) {
1424 f2fs_put_page(page, 1);
1425 return;
1426 }
46706d59 1427
af697c0f 1428 f2fs_bug_on(sbi, err);
46706d59
GX		 1429 f2fs_put_page(page, 0);
1430
1431 /* submit checkpoint (with barrier if NOBARRIER is not set) */
1432 f2fs_submit_merged_write(sbi, META_FLUSH);
1433}
1434
3a0a9cbc
CY		 1435static inline u64 get_sectors_written(struct block_device *bdev)
1436{
ff49c86f 1437 return (u64)part_stat_read(bdev, sectors[STAT_WRITE]);
3a0a9cbc
CY		 1438}
1439
1440u64 f2fs_get_sectors_written(struct f2fs_sb_info *sbi)
1441{
1442 if (f2fs_is_multi_device(sbi)) {
1443 u64 sectors = 0;
1444 int i;
1445
1446 for (i = 0; i < sbi->s_ndevs; i++)
1447 sectors += get_sectors_written(FDEV(i).bdev);
1448
1449 return sectors;
1450 }
1451
1452 return get_sectors_written(sbi->sb->s_bdev);
1453}
1454
c34f42e2 1455static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
127e670a
JK		 1456{
1457 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
77041823 1458 struct f2fs_nm_info *nm_i = NM_I(sbi);
d1aa2453 1459 unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num, flags;
127e670a 1460 block_t start_blk;
127e670a 1461 unsigned int data_sum_blocks, orphan_blocks;
7e586fa0 1462 __u32 crc32 = 0;
127e670a 1463 int i;
55141486 1464 int cp_payload_blks = __cp_payload(sbi);
8f1dbbbb
SL		 1465 struct curseg_info *seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
1466 u64 kbytes_written;
1228b482 1467 int err;
127e670a
JK		 1468
1469 /* Flush all the NAT/SIT pages */
8ec18bff 1470 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO);
127e670a 1471
7a88ddb5 1472 /* start to update checkpoint, cp ver is already updated previously */
a1f72ac2 1473 ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi, true));
127e670a 1474 ckpt->free_segment_count = cpu_to_le32(free_segments(sbi));
b5b82205 1475 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
127e670a
JK		 1476 ckpt->cur_node_segno[i] =
1477 cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_NODE));
1478 ckpt->cur_node_blkoff[i] =
1479 cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_NODE));
1480 ckpt->alloc_type[i + CURSEG_HOT_NODE] =
1481 curseg_alloc_type(sbi, i + CURSEG_HOT_NODE);
1482 }
b5b82205 1483 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
127e670a
JK		 1484 ckpt->cur_data_segno[i] =
1485 cpu_to_le32(curseg_segno(sbi, i + CURSEG_HOT_DATA));
1486 ckpt->cur_data_blkoff[i] =
1487 cpu_to_le16(curseg_blkoff(sbi, i + CURSEG_HOT_DATA));
1488 ckpt->alloc_type[i + CURSEG_HOT_DATA] =
1489 curseg_alloc_type(sbi, i + CURSEG_HOT_DATA);
1490 }
1491
a87aff1d 1492 /* 2 cp + n data seg summary + orphan inode blocks */
4d57b86d 1493 data_sum_blocks = f2fs_npages_for_summary_flush(sbi, false);
d1aa2453 1494 spin_lock_irqsave(&sbi->cp_lock, flags);
b5b82205 1495 if (data_sum_blocks < NR_CURSEG_DATA_TYPE)
aaec2b1d 1496 __set_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG);
127e670a 1497 else
aaec2b1d 1498 __clear_ckpt_flags(ckpt, CP_COMPACT_SUM_FLAG);
d1aa2453 1499 spin_unlock_irqrestore(&sbi->cp_lock, flags);
127e670a 1500
67298804 1501 orphan_blocks = GET_ORPHAN_BLOCKS(orphan_num);
1dbe4152
CL		 1502 ckpt->cp_pack_start_sum = cpu_to_le32(1 + cp_payload_blks +
1503 orphan_blocks);
127e670a 1504
119ee914 1505 if (__remain_node_summaries(cpc->reason))
2a4bd0c3 1506 ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS +
1dbe4152
CL		 1507 cp_payload_blks + data_sum_blocks +
1508 orphan_blocks + NR_CURSEG_NODE_TYPE);
119ee914 1509 else
b5b82205 1510 ckpt->cp_pack_total_block_count = cpu_to_le32(F2FS_CP_PACKS +
1dbe4152
CL		 1511 cp_payload_blks + data_sum_blocks +
1512 orphan_blocks);
119ee914 1513
e4c5d848
JK		 1514 /* update ckpt flag for checkpoint */
1515 update_ckpt_flags(sbi, cpc);
a468f0ef 1516
127e670a
JK		 1517 /* update SIT/NAT bitmap */
1518 get_sit_bitmap(sbi, __bitmap_ptr(sbi, SIT_BITMAP));
1519 get_nat_bitmap(sbi, __bitmap_ptr(sbi, NAT_BITMAP));
1520
d7eb8f1c 1521 crc32 = f2fs_checkpoint_chksum(sbi, ckpt);
7e586fa0
JK		 1522 *((__le32 *)((unsigned char *)ckpt +
1523 le32_to_cpu(ckpt->checksum_offset)))
127e670a
JK		 1524 = cpu_to_le32(crc32);
1525
8508e44a 1526 start_blk = __start_cp_next_addr(sbi);
127e670a 1527
22ad0b6a 1528 /* write nat bits */
94c821fb
CY		 1529 if ((cpc->reason & CP_UMOUNT) &&
1530 is_set_ckpt_flags(sbi, CP_NAT_BITS_FLAG)) {
22ad0b6a 1531 __u64 cp_ver = cur_cp_version(ckpt);
22ad0b6a
JK		 1532 block_t blk;
1533
1534 cp_ver |= ((__u64)crc32 << 32);
1535 *(__le64 *)nm_i->nat_bits = cpu_to_le64(cp_ver);
1536
1537 blk = start_blk + sbi->blocks_per_seg - nm_i->nat_bits_blocks;
1538 for (i = 0; i < nm_i->nat_bits_blocks; i++)
4d57b86d 1539 f2fs_update_meta_page(sbi, nm_i->nat_bits +
22ad0b6a 1540 (i << F2FS_BLKSIZE_BITS), blk + i);
22ad0b6a
JK		 1541 }
1542
127e670a 1543 /* write out checkpoint buffer at block 0 */
4d57b86d 1544 f2fs_update_meta_page(sbi, ckpt, start_blk++);
381722d2
CY		 1545
1546 for (i = 1; i < 1 + cp_payload_blks; i++)
4d57b86d 1547 f2fs_update_meta_page(sbi, (char *)ckpt + i * F2FS_BLKSIZE,
381722d2 1548 start_blk++);
1dbe4152 1549
67298804 1550 if (orphan_num) {
127e670a
JK		 1551 write_orphan_inodes(sbi, start_blk);
1552 start_blk += orphan_blocks;
1553 }
1554
4d57b86d 1555 f2fs_write_data_summaries(sbi, start_blk);
127e670a 1556 start_blk += data_sum_blocks;
8f1dbbbb
SL		 1557
1558 /* Record write statistics in the hot node summary */
1559 kbytes_written = sbi->kbytes_written;
3a0a9cbc
CY		 1560 kbytes_written += (f2fs_get_sectors_written(sbi) -
1561 sbi->sectors_written_start) >> 1;
b7ad7512 1562 seg_i->journal->info.kbytes_written = cpu_to_le64(kbytes_written);
8f1dbbbb 1563
119ee914 1564 if (__remain_node_summaries(cpc->reason)) {
4d57b86d 1565 f2fs_write_node_summaries(sbi, start_blk);
127e670a
JK		 1566 start_blk += NR_CURSEG_NODE_TYPE;
1567 }
1568
46706d59
GX		 1569 /* update user_block_counts */
1570 sbi->last_valid_block_count = sbi->total_valid_block_count;
1571 percpu_counter_set(&sbi->alloc_valid_block_count, 0);
47c8ebcc 1572 percpu_counter_set(&sbi->rf_node_block_count, 0);
127e670a 1573
46706d59 1574 /* Here, we have one bio having CP pack except cp pack 2 page */
4d57b86d 1575 f2fs_sync_meta_pages(sbi, META, LONG_MAX, FS_CP_META_IO);
bf22c3cc
ST		 1576 /* Wait for all dirty meta pages to be submitted for IO */
1577 f2fs_wait_on_all_pages(sbi, F2FS_DIRTY_META);
46706d59
GX		 1578
1579 /* wait for previous submitted meta pages writeback */
bf22c3cc 1580 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
127e670a 1581
46706d59
GX		 1582 /* flush all device cache */
1583 err = f2fs_flush_device_cache(sbi);
1584 if (err)
1585 return err;
127e670a 1586
46706d59
GX		 1587 /* barrier and flush checkpoint cp pack 2 page if it can */
1588 commit_checkpoint(sbi, ckpt, start_blk);
bf22c3cc 1589 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
6a8f8ca5 1590
18767e62 1591 /*
542989b6 1592 * invalidate intermediate page cache borrowed from meta inode which are
aff6fbbe 1593 * used for migration of encrypted, verity or compressed inode's blocks.
18767e62 1594 */
aff6fbbe
CY		 1595 if (f2fs_sb_has_encrypt(sbi) || f2fs_sb_has_verity(sbi) ||
1596 f2fs_sb_has_compression(sbi))
18767e62
CY		 1597 invalidate_mapping_pages(META_MAPPING(sbi),
1598 MAIN_BLKADDR(sbi), MAX_BLKADDR(sbi) - 1);
1599
4d57b86d 1600 f2fs_release_ino_entry(sbi, false);
cf779cab 1601
50fa53ec
CY		 1602 f2fs_reset_fsync_node_info(sbi);
1603
caf0047e 1604 clear_sbi_flag(sbi, SBI_IS_DIRTY);
bbf156f7 1605 clear_sbi_flag(sbi, SBI_NEED_CP);
af033b2a 1606 clear_sbi_flag(sbi, SBI_QUOTA_SKIP_FLUSH);
c9c8ed50
CY		 1607
1608 spin_lock(&sbi->stat_lock);
4354994f 1609 sbi->unusable_block_count = 0;
c9c8ed50
CY		 1610 spin_unlock(&sbi->stat_lock);
1611
8508e44a 1612 __set_cp_next_pack(sbi);
c34f42e2 1613
c2a080ae
CY		 1614 /*
1615 * redirty superblock if metadata like node page or inode cache is
1616 * updated during writing checkpoint.
1617 */
1618 if (get_pages(sbi, F2FS_DIRTY_NODES) ||
1619 get_pages(sbi, F2FS_DIRTY_IMETA))
1620 set_sbi_flag(sbi, SBI_IS_DIRTY);
1621
1622 f2fs_bug_on(sbi, get_pages(sbi, F2FS_DIRTY_DENTS));
1623
af697c0f 1624 return unlikely(f2fs_cp_error(sbi)) ? -EIO : 0;
127e670a
JK		 1625}
1626
4d57b86d 1627int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
127e670a
JK		 1628{
1629 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
1630 unsigned long long ckpt_ver;
c34f42e2 1631 int err = 0;
127e670a 1632
f5a131bb
CY		 1633 if (f2fs_readonly(sbi->sb) || f2fs_hw_is_readonly(sbi))
1634 return -EROFS;
1635
4354994f
DR		 1636 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1637 if (cpc->reason != CP_PAUSE)
1638 return 0;
dcbb4c10 1639 f2fs_warn(sbi, "Start checkpoint disabled!");
4354994f 1640 }
b4b10061 1641 if (cpc->reason != CP_RESIZE)
e4544b63 1642 f2fs_down_write(&sbi->cp_global_sem);
8501017e 1643
caf0047e 1644 if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) &&
c473f1a9
CY		 1645 ((cpc->reason & CP_FASTBOOT) || (cpc->reason & CP_SYNC) ||
1646 ((cpc->reason & CP_DISCARD) && !sbi->discard_blks)))
8501017e 1647 goto out;
c34f42e2
CY		 1648 if (unlikely(f2fs_cp_error(sbi))) {
1649 err = -EIO;
cf779cab 1650 goto out;
c34f42e2 1651 }
2bda542d
WL		 1652
1653 trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops");
1654
c34f42e2
CY		 1655 err = block_operations(sbi);
1656 if (err)
cf779cab 1657 goto out;
127e670a 1658
75ab4cb8 1659 trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish block_ops");
2af4bd6c 1660
b9109b0e 1661 f2fs_flush_merged_writes(sbi);
127e670a 1662
58cce381 1663 /* this is the case of multiple fstrims without any changes */
c473f1a9 1664 if (cpc->reason & CP_DISCARD) {
4d57b86d 1665 if (!f2fs_exist_trim_candidates(sbi, cpc)) {
25290fa5
JK		 1666 unblock_operations(sbi);
1667 goto out;
1668 }
1669
a95ba66a 1670 if (NM_I(sbi)->nat_cnt[DIRTY_NAT] == 0 &&
0333ad4e
JK		 1671 SIT_I(sbi)->dirty_sentries == 0 &&
1672 prefree_segments(sbi) == 0) {
4d57b86d
CY		 1673 f2fs_flush_sit_entries(sbi, cpc);
1674 f2fs_clear_prefree_segments(sbi, cpc);
0333ad4e
JK		 1675 unblock_operations(sbi);
1676 goto out;
1677 }
58cce381
YH		 1678 }
1679
127e670a
JK		 1680 /*
1681 * update checkpoint pack index
1682 * Increase the version number so that
1683 * SIT entries and seg summaries are written at correct place
1684 */
d71b5564 1685 ckpt_ver = cur_cp_version(ckpt);
127e670a
JK		 1686 ckpt->checkpoint_ver = cpu_to_le64(++ckpt_ver);
1687
1688 /* write cached NAT/SIT entries to NAT/SIT area */
edc55aaf 1689 err = f2fs_flush_nat_entries(sbi, cpc);
91803392
CY		 1690 if (err) {
1691 f2fs_err(sbi, "f2fs_flush_nat_entries failed err:%d, stop checkpoint", err);
1692 f2fs_bug_on(sbi, !f2fs_cp_error(sbi));
edc55aaf 1693 goto stop;
91803392 1694 }
edc55aaf 1695
4d57b86d 1696 f2fs_flush_sit_entries(sbi, cpc);
127e670a 1697
d0b9e42a 1698 /* save inmem log status */
093749e2 1699 f2fs_save_inmem_curseg(sbi);
d0b9e42a 1700
c34f42e2 1701 err = do_checkpoint(sbi, cpc);
91803392
CY		 1702 if (err) {
1703 f2fs_err(sbi, "do_checkpoint failed err:%d, stop checkpoint", err);
1704 f2fs_bug_on(sbi, !f2fs_cp_error(sbi));
4d57b86d 1705 f2fs_release_discard_addrs(sbi);
91803392 1706 } else {
4d57b86d 1707 f2fs_clear_prefree_segments(sbi, cpc);
91803392 1708 }
d0b9e42a 1709
093749e2 1710 f2fs_restore_inmem_curseg(sbi);
edc55aaf 1711stop:
127e670a 1712 unblock_operations(sbi);
942e0be6 1713 stat_inc_cp_count(sbi->stat_info);
10027551 1714
c473f1a9 1715 if (cpc->reason & CP_RECOVERY)
dcbb4c10 1716 f2fs_notice(sbi, "checkpoint: version = %llx", ckpt_ver);
60b99b48 1717
7a88ddb5 1718 /* update CP_TIME to trigger checkpoint periodically */
6beceb54 1719 f2fs_update_time(sbi, CP_TIME);
55d1cdb2 1720 trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint");
8501017e 1721out:
b4b10061 1722 if (cpc->reason != CP_RESIZE)
e4544b63 1723 f2fs_up_write(&sbi->cp_global_sem);
c34f42e2 1724 return err;
127e670a
JK		 1725}
1726
4d57b86d 1727void f2fs_init_ino_entry_info(struct f2fs_sb_info *sbi)
127e670a 1728{
6451e041
JK		 1729 int i;
1730
1731 for (i = 0; i < MAX_INO_ENTRY; i++) {
67298804
CY		 1732 struct inode_management *im = &sbi->im[i];
1733
1734 INIT_RADIX_TREE(&im->ino_root, GFP_ATOMIC);
1735 spin_lock_init(&im->ino_lock);
1736 INIT_LIST_HEAD(&im->ino_list);
1737 im->ino_num = 0;
6451e041
JK		 1738 }
1739
b5b82205 1740 sbi->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS -
d0b9e42a 1741 NR_CURSEG_PERSIST_TYPE - __cp_payload(sbi)) *
14b42817 1742 F2FS_ORPHANS_PER_BLOCK;
127e670a
JK		 1743}
1744
4d57b86d 1745int __init f2fs_create_checkpoint_caches(void)
127e670a 1746{
6451e041
JK		 1747 ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry",
1748 sizeof(struct ino_entry));
1749 if (!ino_entry_slab)
127e670a 1750 return -ENOMEM;
4d57b86d 1751 f2fs_inode_entry_slab = f2fs_kmem_cache_create("f2fs_inode_entry",
06292073 1752 sizeof(struct inode_entry));
4d57b86d 1753 if (!f2fs_inode_entry_slab) {
6451e041 1754 kmem_cache_destroy(ino_entry_slab);
127e670a
JK		 1755 return -ENOMEM;
1756 }
1757 return 0;
1758}
1759
4d57b86d 1760void f2fs_destroy_checkpoint_caches(void)
127e670a 1761{
6451e041 1762 kmem_cache_destroy(ino_entry_slab);
4d57b86d 1763 kmem_cache_destroy(f2fs_inode_entry_slab);
127e670a 1764}
261eeb9c
DJ		 1765
1766static int __write_checkpoint_sync(struct f2fs_sb_info *sbi)
1767{
1768 struct cp_control cpc = { .reason = CP_SYNC, };
1769 int err;
1770
e4544b63 1771 f2fs_down_write(&sbi->gc_lock);
261eeb9c 1772 err = f2fs_write_checkpoint(sbi, &cpc);
e4544b63 1773 f2fs_up_write(&sbi->gc_lock);
261eeb9c
DJ		 1774
1775 return err;
1776}
1777
1778static void __checkpoint_and_complete_reqs(struct f2fs_sb_info *sbi)
1779{
1780 struct ckpt_req_control *cprc = &sbi->cprc_info;
1781 struct ckpt_req *req, *next;
1782 struct llist_node *dispatch_list;
1783 u64 sum_diff = 0, diff, count = 0;
1784 int ret;
1785
1786 dispatch_list = llist_del_all(&cprc->issue_list);
1787 if (!dispatch_list)
1788 return;
1789 dispatch_list = llist_reverse_order(dispatch_list);
1790
1791 ret = __write_checkpoint_sync(sbi);
1792 atomic_inc(&cprc->issued_ckpt);
1793
1794 llist_for_each_entry_safe(req, next, dispatch_list, llnode) {
1795 diff = (u64)ktime_ms_delta(ktime_get(), req->queue_time);
1796 req->ret = ret;
1797 complete(&req->wait);
1798
1799 sum_diff += diff;
1800 count++;
1801 }
1802 atomic_sub(count, &cprc->queued_ckpt);
1803 atomic_add(count, &cprc->total_ckpt);
1804
1805 spin_lock(&cprc->stat_lock);
1806 cprc->cur_time = (unsigned int)div64_u64(sum_diff, count);
1807 if (cprc->peak_time < cprc->cur_time)
1808 cprc->peak_time = cprc->cur_time;
1809 spin_unlock(&cprc->stat_lock);
1810}
1811
1812static int issue_checkpoint_thread(void *data)
1813{
1814 struct f2fs_sb_info *sbi = data;
1815 struct ckpt_req_control *cprc = &sbi->cprc_info;
1816 wait_queue_head_t *q = &cprc->ckpt_wait_queue;
1817repeat:
1818 if (kthread_should_stop())
1819 return 0;
1820
261eeb9c
DJ		 1821 if (!llist_empty(&cprc->issue_list))
1822 __checkpoint_and_complete_reqs(sbi);
1823
261eeb9c
DJ		 1824 wait_event_interruptible(*q,
1825 kthread_should_stop() || !llist_empty(&cprc->issue_list));
1826 goto repeat;
1827}
1828
1829static void flush_remained_ckpt_reqs(struct f2fs_sb_info *sbi,
1830 struct ckpt_req *wait_req)
1831{
1832 struct ckpt_req_control *cprc = &sbi->cprc_info;
1833
1834 if (!llist_empty(&cprc->issue_list)) {
1835 __checkpoint_and_complete_reqs(sbi);
1836 } else {
1837 /* already dispatched by issue_checkpoint_thread */
1838 if (wait_req)
1839 wait_for_completion(&wait_req->wait);
1840 }
1841}
1842
1843static void init_ckpt_req(struct ckpt_req *req)
1844{
1845 memset(req, 0, sizeof(struct ckpt_req));
1846
1847 init_completion(&req->wait);
1848 req->queue_time = ktime_get();
1849}
1850
1851int f2fs_issue_checkpoint(struct f2fs_sb_info *sbi)
1852{
1853 struct ckpt_req_control *cprc = &sbi->cprc_info;
1854 struct ckpt_req req;
1855 struct cp_control cpc;
1856
1857 cpc.reason = __get_cp_reason(sbi);
1858 if (!test_opt(sbi, MERGE_CHECKPOINT) || cpc.reason != CP_SYNC) {
1859 int ret;
1860
e4544b63 1861 f2fs_down_write(&sbi->gc_lock);
261eeb9c 1862 ret = f2fs_write_checkpoint(sbi, &cpc);
e4544b63 1863 f2fs_up_write(&sbi->gc_lock);
261eeb9c
DJ		 1864
1865 return ret;
1866 }
1867
1868 if (!cprc->f2fs_issue_ckpt)
1869 return __write_checkpoint_sync(sbi);
1870
1871 init_ckpt_req(&req);
1872
1873 llist_add(&req.llnode, &cprc->issue_list);
1874 atomic_inc(&cprc->queued_ckpt);
1875
3b42c741
CY		 1876 /*
1877 * update issue_list before we wake up issue_checkpoint thread,
1878 * this smp_mb() pairs with another barrier in ___wait_event(),
1879 * see more details in comments of waitqueue_active().
1880 */
261eeb9c
DJ		 1881 smp_mb();
1882
1883 if (waitqueue_active(&cprc->ckpt_wait_queue))
1884 wake_up(&cprc->ckpt_wait_queue);
1885
1886 if (cprc->f2fs_issue_ckpt)
1887 wait_for_completion(&req.wait);
1888 else
1889 flush_remained_ckpt_reqs(sbi, &req);
1890
1891 return req.ret;
1892}
1893
1894int f2fs_start_ckpt_thread(struct f2fs_sb_info *sbi)
1895{
1896 dev_t dev = sbi->sb->s_bdev->bd_dev;
1897 struct ckpt_req_control *cprc = &sbi->cprc_info;
1898
1899 if (cprc->f2fs_issue_ckpt)
1900 return 0;
1901
1902 cprc->f2fs_issue_ckpt = kthread_run(issue_checkpoint_thread, sbi,
1903 "f2fs_ckpt-%u:%u", MAJOR(dev), MINOR(dev));
1904 if (IS_ERR(cprc->f2fs_issue_ckpt)) {
146dbcbf
YL		 1905 int err = PTR_ERR(cprc->f2fs_issue_ckpt);
1906
261eeb9c 1907 cprc->f2fs_issue_ckpt = NULL;
146dbcbf 1908 return err;
261eeb9c
DJ		 1909 }
1910
e6592066 1911 set_task_ioprio(cprc->f2fs_issue_ckpt, cprc->ckpt_thread_ioprio);
261eeb9c
DJ		 1912
1913 return 0;
1914}
1915
1916void f2fs_stop_ckpt_thread(struct f2fs_sb_info *sbi)
1917{
1918 struct ckpt_req_control *cprc = &sbi->cprc_info;
c7b58576 1919 struct task_struct *ckpt_task;
261eeb9c 1920
c7b58576
JK		 1921 if (!cprc->f2fs_issue_ckpt)
1922 return;
261eeb9c 1923
c7b58576
JK		 1924 ckpt_task = cprc->f2fs_issue_ckpt;
1925 cprc->f2fs_issue_ckpt = NULL;
1926 kthread_stop(ckpt_task);
261eeb9c 1927
c7b58576
JK		 1928 f2fs_flush_ckpt_thread(sbi);
1929}
1930
1931void f2fs_flush_ckpt_thread(struct f2fs_sb_info *sbi)
1932{
1933 struct ckpt_req_control *cprc = &sbi->cprc_info;
1934
1935 flush_remained_ckpt_reqs(sbi, NULL);
1936
1937 /* Let's wait for the previous dispatched checkpoint. */
1938 while (atomic_read(&cprc->queued_ckpt))
1939 io_schedule_timeout(DEFAULT_IO_TIMEOUT);
261eeb9c
DJ		 1940}
1941
1942void f2fs_init_ckpt_req_control(struct f2fs_sb_info *sbi)
1943{
1944 struct ckpt_req_control *cprc = &sbi->cprc_info;
1945
1946 atomic_set(&cprc->issued_ckpt, 0);
1947 atomic_set(&cprc->total_ckpt, 0);
1948 atomic_set(&cprc->queued_ckpt, 0);
e6592066 1949 cprc->ckpt_thread_ioprio = DEFAULT_CHECKPOINT_IOPRIO;
261eeb9c
DJ		 1950 init_waitqueue_head(&cprc->ckpt_wait_queue);
1951 init_llist_head(&cprc->issue_list);
1952 spin_lock_init(&cprc->stat_lock);
1953}
Linux 6.x block layer and io_uring tree(s)