projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
f2fs: clean up commit_inmem_pages()
[linux-2.6-block.git] / fs / f2fs / segment.h
CommitLineData
0a8165d7 1/*
39a53e0c
JK		 2 * fs/f2fs/segment.h
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
ac5d156c 11#include <linux/blkdev.h>
66114cad 12#include <linux/backing-dev.h>
ac5d156c 13
39a53e0c
JK		 14/* constant macro */
15#define NULL_SEGNO ((unsigned int)(~0))
5ec4e49f 16#define NULL_SECNO ((unsigned int)(~0))
39a53e0c 17
58c41035 18#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
44a83499 19#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
81eb8d6e 20
2040fce8
JK		 21#define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
22
6224da87 23/* L: Logical segment # in volume, R: Relative segment # in main area */
68afcf2d
TK		 24#define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
25#define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
39a53e0c 26
68afcf2d
TK		 27#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
28#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
39a53e0c 29
a912b54d
JK		 30#define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
31#define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
32#define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
33
5c773ba3 34#define IS_CURSEG(sbi, seg) \
68afcf2d
TK		 35 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
36 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
37 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
38 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
39 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
40 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
39a53e0c
JK		 41
42#define IS_CURSEC(sbi, secno) \
68afcf2d
TK		 43 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
44 (sbi)->segs_per_sec) || \
45 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
46 (sbi)->segs_per_sec) || \
47 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
48 (sbi)->segs_per_sec) || \
49 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
50 (sbi)->segs_per_sec) || \
51 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
52 (sbi)->segs_per_sec) || \
53 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
54 (sbi)->segs_per_sec)) \
39a53e0c 55
0833721e
YH		 56#define MAIN_BLKADDR(sbi) \
57 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
58 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
59#define SEG0_BLKADDR(sbi) \
60 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
61 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
7cd8558b
JK		 62
63#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
68afcf2d 64#define MAIN_SECS(sbi) ((sbi)->total_sections)
7cd8558b 65
0833721e
YH		 66#define TOTAL_SEGS(sbi) \
67 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
68 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
68afcf2d 69#define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
7cd8558b
JK		 70
71#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
68afcf2d
TK		 72#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
73 (sbi)->log_blocks_per_seg))
7cd8558b
JK		 74
75#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
68afcf2d 76 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
7cd8558b 77
39a53e0c 78#define NEXT_FREE_BLKADDR(sbi, curseg) \
68afcf2d 79 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
39a53e0c 80
7cd8558b 81#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
39a53e0c 82#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
68afcf2d 83 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
491c0854 84#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
68afcf2d 85 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
491c0854 86
39a53e0c 87#define GET_SEGNO(sbi, blk_addr) \
68afcf2d 88 ((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ? \
39a53e0c
JK		 89 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
90 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
4ddb1a4d
JK		 91#define BLKS_PER_SEC(sbi) \
92 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
93#define GET_SEC_FROM_SEG(sbi, segno) \
68afcf2d 94 ((segno) / (sbi)->segs_per_sec)
4ddb1a4d 95#define GET_SEG_FROM_SEC(sbi, secno) \
63fcf8e8 96 ((secno) * (sbi)->segs_per_sec)
4ddb1a4d
JK		 97#define GET_ZONE_FROM_SEC(sbi, secno) \
98 ((secno) / (sbi)->secs_per_zone)
99#define GET_ZONE_FROM_SEG(sbi, segno) \
100 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
39a53e0c
JK		 101
102#define GET_SUM_BLOCK(sbi, segno) \
68afcf2d 103 ((sbi)->sm_info->ssa_blkaddr + (segno))
39a53e0c
JK		 104
105#define GET_SUM_TYPE(footer) ((footer)->entry_type)
68afcf2d 106#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
39a53e0c
JK		 107
108#define SIT_ENTRY_OFFSET(sit_i, segno) \
68afcf2d 109 ((segno) % (sit_i)->sents_per_block)
d3a14afd 110#define SIT_BLOCK_OFFSET(segno) \
68afcf2d 111 ((segno) / SIT_ENTRY_PER_BLOCK)
d3a14afd
CY		 112#define START_SEGNO(segno) \
113 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
74de593a 114#define SIT_BLK_CNT(sbi) \
7cd8558b 115 ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
39a53e0c
JK		 116#define f2fs_bitmap_size(nr) \
117 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
39a53e0c 118
55cf9cb6
CY		 119#define SECTOR_FROM_BLOCK(blk_addr) \
120 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
121#define SECTOR_TO_BLOCK(sectors) \
68afcf2d 122 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
3cd8a239 123
39a53e0c
JK		 124/*
125 * indicate a block allocation direction: RIGHT and LEFT.
126 * RIGHT means allocating new sections towards the end of volume.
127 * LEFT means the opposite direction.
128 */
129enum {
130 ALLOC_RIGHT = 0,
131 ALLOC_LEFT
132};
133
134/*
135 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
136 * LFS writes data sequentially with cleaning operations.
137 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
138 */
139enum {
140 LFS = 0,
141 SSR
142};
143
144/*
145 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
146 * GC_CB is based on cost-benefit algorithm.
147 * GC_GREEDY is based on greedy algorithm.
148 */
149enum {
150 GC_CB = 0,
e066b83c
JK		 151 GC_GREEDY,
152 ALLOC_NEXT,
153 FLUSH_DEVICE,
154 MAX_GC_POLICY,
39a53e0c
JK		 155};
156
157/*
158 * BG_GC means the background cleaning job.
159 * FG_GC means the on-demand cleaning job.
6aefd93b 160 * FORCE_FG_GC means on-demand cleaning job in background.
39a53e0c
JK		 161 */
162enum {
163 BG_GC = 0,
6aefd93b
JK		 164 FG_GC,
165 FORCE_FG_GC,
39a53e0c
JK		 166};
167
168/* for a function parameter to select a victim segment */
169struct victim_sel_policy {
170 int alloc_mode; /* LFS or SSR */
171 int gc_mode; /* GC_CB or GC_GREEDY */
172 unsigned long *dirty_segmap; /* dirty segment bitmap */
a26b7c8a 173 unsigned int max_search; /* maximum # of segments to search */
39a53e0c
JK		 174 unsigned int offset; /* last scanned bitmap offset */
175 unsigned int ofs_unit; /* bitmap search unit */
176 unsigned int min_cost; /* minimum cost */
177 unsigned int min_segno; /* segment # having min. cost */
178};
179
180struct seg_entry {
f51b4ce6
CY		 181 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
182 unsigned int valid_blocks:10; /* # of valid blocks */
183 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
184 unsigned int padding:6; /* padding */
39a53e0c 185 unsigned char *cur_valid_map; /* validity bitmap of blocks */
355e7891
CY		 186#ifdef CONFIG_F2FS_CHECK_FS
187 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
188#endif
39a53e0c
JK		 189 /*
190 * # of valid blocks and the validity bitmap stored in the the last
191 * checkpoint pack. This information is used by the SSR mode.
192 */
f51b4ce6 193 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
a66cdd98 194 unsigned char *discard_map;
39a53e0c
JK		 195 unsigned long long mtime; /* modification time of the segment */
196};
197
198struct sec_entry {
199 unsigned int valid_blocks; /* # of valid blocks in a section */
200};
201
202struct segment_allocation {
203 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
204};
205
decd36b6
CY		 206/*
207 * this value is set in page as a private data which indicate that
208 * the page is atomically written, and it is in inmem_pages list.
209 */
d48dfc20 210#define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
0a595eba 211#define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
decd36b6
CY		 212
213#define IS_ATOMIC_WRITTEN_PAGE(page) \
214 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
0a595eba
JK		 215#define IS_DUMMY_WRITTEN_PAGE(page) \
216 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
decd36b6 217
88b88a66
JK		 218struct inmem_pages {
219 struct list_head list;
220 struct page *page;
28bc106b 221 block_t old_addr; /* for revoking when fail to commit */
88b88a66
JK		 222};
223
39a53e0c
JK		 224struct sit_info {
225 const struct segment_allocation *s_ops;
226
227 block_t sit_base_addr; /* start block address of SIT area */
228 block_t sit_blocks; /* # of blocks used by SIT area */
229 block_t written_valid_blocks; /* # of valid blocks in main area */
230 char *sit_bitmap; /* SIT bitmap pointer */
ae27d62e
CY		 231#ifdef CONFIG_F2FS_CHECK_FS
232 char *sit_bitmap_mir; /* SIT bitmap mirror */
233#endif
39a53e0c
JK		 234 unsigned int bitmap_size; /* SIT bitmap size */
235
60a3b782 236 unsigned long *tmp_map; /* bitmap for temporal use */
39a53e0c
JK		 237 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
238 unsigned int dirty_sentries; /* # of dirty sentries */
239 unsigned int sents_per_block; /* # of SIT entries per block */
3d26fa6b 240 struct rw_semaphore sentry_lock; /* to protect SIT cache */
39a53e0c
JK		 241 struct seg_entry *sentries; /* SIT segment-level cache */
242 struct sec_entry *sec_entries; /* SIT section-level cache */
243
244 /* for cost-benefit algorithm in cleaning procedure */
245 unsigned long long elapsed_time; /* elapsed time after mount */
246 unsigned long long mounted_time; /* mount time */
247 unsigned long long min_mtime; /* min. modification time */
248 unsigned long long max_mtime; /* max. modification time */
e066b83c
JK		 249
250 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
39a53e0c
JK		 251};
252
253struct free_segmap_info {
254 unsigned int start_segno; /* start segment number logically */
255 unsigned int free_segments; /* # of free segments */
256 unsigned int free_sections; /* # of free sections */
1a118ccf 257 spinlock_t segmap_lock; /* free segmap lock */
39a53e0c
JK		 258 unsigned long *free_segmap; /* free segment bitmap */
259 unsigned long *free_secmap; /* free section bitmap */
260};
261
262/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
263enum dirty_type {
264 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
265 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
266 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
267 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
268 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
269 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
270 DIRTY, /* to count # of dirty segments */
271 PRE, /* to count # of entirely obsolete segments */
272 NR_DIRTY_TYPE
273};
274
275struct dirty_seglist_info {
276 const struct victim_selection *v_ops; /* victim selction operation */
277 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
278 struct mutex seglist_lock; /* lock for segment bitmaps */
279 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
5ec4e49f 280 unsigned long *victim_secmap; /* background GC victims */
39a53e0c
JK		 281};
282
283/* victim selection function for cleaning and SSR */
284struct victim_selection {
285 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
286 int, int, char);
287};
288
289/* for active log information */
290struct curseg_info {
291 struct mutex curseg_mutex; /* lock for consistency */
292 struct f2fs_summary_block *sum_blk; /* cached summary block */
b7ad7512
CY		 293 struct rw_semaphore journal_rwsem; /* protect journal area */
294 struct f2fs_journal *journal; /* cached journal info */
39a53e0c
JK		 295 unsigned char alloc_type; /* current allocation type */
296 unsigned int segno; /* current segment number */
297 unsigned short next_blkoff; /* next block offset to write */
298 unsigned int zone; /* current zone number */
299 unsigned int next_segno; /* preallocated segment */
300};
301
184a5cd2
CY		 302struct sit_entry_set {
303 struct list_head set_list; /* link with all sit sets */
304 unsigned int start_segno; /* start segno of sits in set */
305 unsigned int entry_cnt; /* the # of sit entries in set */
306};
307
39a53e0c
JK		 308/*
309 * inline functions
310 */
311static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
312{
313 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
314}
315
316static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
317 unsigned int segno)
318{
319 struct sit_info *sit_i = SIT_I(sbi);
320 return &sit_i->sentries[segno];
321}
322
323static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
324 unsigned int segno)
325{
326 struct sit_info *sit_i = SIT_I(sbi);
4ddb1a4d 327 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
39a53e0c
JK		 328}
329
330static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
302bd348 331 unsigned int segno, bool use_section)
39a53e0c
JK		 332{
333 /*
334 * In order to get # of valid blocks in a section instantly from many
335 * segments, f2fs manages two counting structures separately.
336 */
302bd348 337 if (use_section && sbi->segs_per_sec > 1)
39a53e0c
JK		 338 return get_sec_entry(sbi, segno)->valid_blocks;
339 else
340 return get_seg_entry(sbi, segno)->valid_blocks;
341}
342
343static inline void seg_info_from_raw_sit(struct seg_entry *se,
344 struct f2fs_sit_entry *rs)
345{
346 se->valid_blocks = GET_SIT_VBLOCKS(rs);
347 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
348 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
349 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
355e7891
CY		 350#ifdef CONFIG_F2FS_CHECK_FS
351 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
352#endif
39a53e0c
JK		 353 se->type = GET_SIT_TYPE(rs);
354 se->mtime = le64_to_cpu(rs->mtime);
355}
356
068c3cd8 357static inline void __seg_info_to_raw_sit(struct seg_entry *se,
39a53e0c
JK		 358 struct f2fs_sit_entry *rs)
359{
360 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
361 se->valid_blocks;
362 rs->vblocks = cpu_to_le16(raw_vblocks);
363 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
068c3cd8
YH		 364 rs->mtime = cpu_to_le64(se->mtime);
365}
366
367static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
368 struct page *page, unsigned int start)
369{
370 struct f2fs_sit_block *raw_sit;
371 struct seg_entry *se;
372 struct f2fs_sit_entry *rs;
373 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
374 (unsigned long)MAIN_SEGS(sbi));
375 int i;
376
377 raw_sit = (struct f2fs_sit_block *)page_address(page);
81114baa 378 memset(raw_sit, 0, PAGE_SIZE);
068c3cd8
YH		 379 for (i = 0; i < end - start; i++) {
380 rs = &raw_sit->entries[i];
381 se = get_seg_entry(sbi, start + i);
382 __seg_info_to_raw_sit(se, rs);
383 }
384}
385
386static inline void seg_info_to_raw_sit(struct seg_entry *se,
387 struct f2fs_sit_entry *rs)
388{
389 __seg_info_to_raw_sit(se, rs);
390
39a53e0c
JK		 391 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
392 se->ckpt_valid_blocks = se->valid_blocks;
39a53e0c
JK		 393}
394
395static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
396 unsigned int max, unsigned int segno)
397{
398 unsigned int ret;
1a118ccf 399 spin_lock(&free_i->segmap_lock);
39a53e0c 400 ret = find_next_bit(free_i->free_segmap, max, segno);
1a118ccf 401 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 402 return ret;
403}
404
405static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
406{
407 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 408 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
409 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK		 410 unsigned int next;
411
1a118ccf 412 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 413 clear_bit(segno, free_i->free_segmap);
414 free_i->free_segments++;
415
7fd97019
WL		 416 next = find_next_bit(free_i->free_segmap,
417 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK		 418 if (next >= start_segno + sbi->segs_per_sec) {
419 clear_bit(secno, free_i->free_secmap);
420 free_i->free_sections++;
421 }
1a118ccf 422 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 423}
424
425static inline void __set_inuse(struct f2fs_sb_info *sbi,
426 unsigned int segno)
427{
428 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 429 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
430
39a53e0c
JK		 431 set_bit(segno, free_i->free_segmap);
432 free_i->free_segments--;
433 if (!test_and_set_bit(secno, free_i->free_secmap))
434 free_i->free_sections--;
435}
436
437static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
438 unsigned int segno)
439{
440 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 441 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
442 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK		 443 unsigned int next;
444
1a118ccf 445 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 446 if (test_and_clear_bit(segno, free_i->free_segmap)) {
447 free_i->free_segments++;
448
f1121ab0
CY		 449 next = find_next_bit(free_i->free_segmap,
450 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK		 451 if (next >= start_segno + sbi->segs_per_sec) {
452 if (test_and_clear_bit(secno, free_i->free_secmap))
453 free_i->free_sections++;
454 }
455 }
1a118ccf 456 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 457}
458
459static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
460 unsigned int segno)
461{
462 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 463 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
464
1a118ccf 465 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 466 if (!test_and_set_bit(segno, free_i->free_segmap)) {
467 free_i->free_segments--;
468 if (!test_and_set_bit(secno, free_i->free_secmap))
469 free_i->free_sections--;
470 }
1a118ccf 471 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 472}
473
474static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
475 void *dst_addr)
476{
477 struct sit_info *sit_i = SIT_I(sbi);
ae27d62e
CY		 478
479#ifdef CONFIG_F2FS_CHECK_FS
480 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
481 sit_i->bitmap_size))
482 f2fs_bug_on(sbi, 1);
483#endif
39a53e0c
JK		 484 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
485}
486
487static inline block_t written_block_count(struct f2fs_sb_info *sbi)
488{
8b8343fa 489 return SIT_I(sbi)->written_valid_blocks;
39a53e0c
JK		 490}
491
492static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
493{
8b8343fa 494 return FREE_I(sbi)->free_segments;
39a53e0c
JK		 495}
496
497static inline int reserved_segments(struct f2fs_sb_info *sbi)
498{
499 return SM_I(sbi)->reserved_segments;
500}
501
502static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
503{
8b8343fa 504 return FREE_I(sbi)->free_sections;
39a53e0c
JK		 505}
506
507static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
508{
509 return DIRTY_I(sbi)->nr_dirty[PRE];
510}
511
512static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
513{
514 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
515 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
516 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
517 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
518 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
519 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
520}
521
522static inline int overprovision_segments(struct f2fs_sb_info *sbi)
523{
524 return SM_I(sbi)->ovp_segments;
525}
526
39a53e0c
JK		 527static inline int reserved_sections(struct f2fs_sb_info *sbi)
528{
4ddb1a4d 529 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
39a53e0c
JK		 530}
531
bf34c93d
CY		 532static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
533{
534 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
535 get_pages(sbi, F2FS_DIRTY_DENTS);
536 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
537 unsigned int segno, left_blocks;
538 int i;
539
540 /* check current node segment */
541 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
542 segno = CURSEG_I(sbi, i)->segno;
543 left_blocks = sbi->blocks_per_seg -
544 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
545
546 if (node_blocks > left_blocks)
547 return false;
548 }
549
550 /* check current data segment */
551 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
552 left_blocks = sbi->blocks_per_seg -
553 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
554 if (dent_blocks > left_blocks)
555 return false;
556 return true;
557}
558
7f3037a5
JK		 559static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
560 int freed, int needed)
39a53e0c 561{
5ac206cf
NJ		 562 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
563 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
b9610bdf 564 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
0f18b462 565
caf0047e 566 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
029cd28c
JK		 567 return false;
568
bf34c93d
CY		 569 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
570 has_curseg_enough_space(sbi))
571 return false;
7f3037a5 572 return (free_sections(sbi) + freed) <=
b9610bdf
JK		 573 (node_secs + 2 * dent_secs + imeta_secs +
574 reserved_sections(sbi) + needed);
39a53e0c
JK		 575}
576
81eb8d6e
JK		 577static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
578{
6c311ec6 579 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
81eb8d6e
JK		 580}
581
39a53e0c
JK		 582static inline int utilization(struct f2fs_sb_info *sbi)
583{
6c311ec6
CF		 584 return div_u64((u64)valid_user_blocks(sbi) * 100,
585 sbi->user_block_count);
39a53e0c
JK		 586}
587
588/*
589 * Sometimes f2fs may be better to drop out-of-place update policy.
216fbd64
JK		 590 * And, users can control the policy through sysfs entries.
591 * There are five policies with triggering conditions as follows.
592 * F2FS_IPU_FORCE - all the time,
593 * F2FS_IPU_SSR - if SSR mode is activated,
594 * F2FS_IPU_UTIL - if FS utilization is over threashold,
595 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
596 * threashold,
c1ce1b02
JK		 597 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
598 * storages. IPU will be triggered only if the # of dirty
599 * pages over min_fsync_blocks.
216fbd64 600 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
39a53e0c 601 */
216fbd64 602#define DEF_MIN_IPU_UTIL 70
c1ce1b02 603#define DEF_MIN_FSYNC_BLOCKS 8
ef095d19 604#define DEF_MIN_HOT_BLOCKS 16
216fbd64 605
84b89e5d
JK		 606#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
607
216fbd64
JK		 608enum {
609 F2FS_IPU_FORCE,
610 F2FS_IPU_SSR,
611 F2FS_IPU_UTIL,
612 F2FS_IPU_SSR_UTIL,
c1ce1b02 613 F2FS_IPU_FSYNC,
04485987 614 F2FS_IPU_ASYNC,
216fbd64
JK		 615};
616
39a53e0c
JK		 617static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
618 int type)
619{
620 struct curseg_info *curseg = CURSEG_I(sbi, type);
621 return curseg->segno;
622}
623
624static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
625 int type)
626{
627 struct curseg_info *curseg = CURSEG_I(sbi, type);
628 return curseg->alloc_type;
629}
630
631static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
632{
633 struct curseg_info *curseg = CURSEG_I(sbi, type);
634 return curseg->next_blkoff;
635}
636
637static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
638{
7a04f64d 639 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
39a53e0c
JK		 640}
641
0833721e 642static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
39a53e0c 643{
0833721e
YH		 644 struct f2fs_sb_info *sbi = fio->sbi;
645
646 if (PAGE_TYPE_OF_BIO(fio->type) == META &&
647 (!is_read_io(fio->op) || fio->is_meta))
648 BUG_ON(blk_addr < SEG0_BLKADDR(sbi) ||
649 blk_addr >= MAIN_BLKADDR(sbi));
650 else
651 BUG_ON(blk_addr < MAIN_BLKADDR(sbi) ||
652 blk_addr >= MAX_BLKADDR(sbi));
39a53e0c
JK		 653}
654
655/*
e1c42045 656 * Summary block is always treated as an invalid block
39a53e0c 657 */
c39a1b34 658static inline int check_block_count(struct f2fs_sb_info *sbi,
39a53e0c
JK		 659 int segno, struct f2fs_sit_entry *raw_sit)
660{
4c278394 661#ifdef CONFIG_F2FS_CHECK_FS
44c60bf2 662 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
39a53e0c 663 int valid_blocks = 0;
44c60bf2 664 int cur_pos = 0, next_pos;
39a53e0c 665
39a53e0c 666 /* check bitmap with valid block count */
44c60bf2
CY		 667 do {
668 if (is_valid) {
669 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
670 sbi->blocks_per_seg,
671 cur_pos);
672 valid_blocks += next_pos - cur_pos;
673 } else
674 next_pos = find_next_bit_le(&raw_sit->valid_map,
675 sbi->blocks_per_seg,
676 cur_pos);
677 cur_pos = next_pos;
678 is_valid = !is_valid;
679 } while (cur_pos < sbi->blocks_per_seg);
c39a1b34
JK		 680
681 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
682 f2fs_msg(sbi->sb, KERN_ERR,
683 "Mismatch valid blocks %d vs. %d",
684 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
685 set_sbi_flag(sbi, SBI_NEED_FSCK);
686 return -EINVAL;
687 }
5d56b671 688#endif
4c278394 689 /* check segment usage, and check boundary of a given segment number */
c39a1b34
JK		 690 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
691 || segno > TOTAL_SEGS(sbi) - 1)) {
692 f2fs_msg(sbi->sb, KERN_ERR,
693 "Wrong valid blocks %d or segno %u",
694 GET_SIT_VBLOCKS(raw_sit), segno);
695 set_sbi_flag(sbi, SBI_NEED_FSCK);
696 return -EINVAL;
697 }
698 return 0;
7a04f64d 699}
39a53e0c
JK		 700
701static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
702 unsigned int start)
703{
704 struct sit_info *sit_i = SIT_I(sbi);
d3a14afd 705 unsigned int offset = SIT_BLOCK_OFFSET(start);
39a53e0c
JK		 706 block_t blk_addr = sit_i->sit_base_addr + offset;
707
708 check_seg_range(sbi, start);
709
ae27d62e
CY		 710#ifdef CONFIG_F2FS_CHECK_FS
711 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
712 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
713 f2fs_bug_on(sbi, 1);
714#endif
715
39a53e0c
JK		 716 /* calculate sit block address */
717 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
718 blk_addr += sit_i->sit_blocks;
719
720 return blk_addr;
721}
722
723static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
724 pgoff_t block_addr)
725{
726 struct sit_info *sit_i = SIT_I(sbi);
727 block_addr -= sit_i->sit_base_addr;
728 if (block_addr < sit_i->sit_blocks)
729 block_addr += sit_i->sit_blocks;
730 else
731 block_addr -= sit_i->sit_blocks;
732
733 return block_addr + sit_i->sit_base_addr;
734}
735
736static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
737{
d3a14afd 738 unsigned int block_off = SIT_BLOCK_OFFSET(start);
39a53e0c 739
c6ac4c0e 740 f2fs_change_bit(block_off, sit_i->sit_bitmap);
ae27d62e
CY		 741#ifdef CONFIG_F2FS_CHECK_FS
742 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
743#endif
39a53e0c
JK		 744}
745
746static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
747{
748 struct sit_info *sit_i = SIT_I(sbi);
48fbfe50
DD		 749 time64_t now = ktime_get_real_seconds();
750
751 return sit_i->elapsed_time + now - sit_i->mounted_time;
39a53e0c
JK		 752}
753
754static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
755 unsigned int ofs_in_node, unsigned char version)
756{
757 sum->nid = cpu_to_le32(nid);
758 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
759 sum->version = version;
760}
761
762static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
763{
764 return __start_cp_addr(sbi) +
765 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
766}
767
768static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
769{
770 return __start_cp_addr(sbi) +
771 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
772 - (base + 1) + type;
773}
5ec4e49f 774
e93b9865
HP		 775static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
776 unsigned int secno)
777{
7e515b31 778 if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >
302bd348 779 sbi->fggc_threshold)
e93b9865
HP		 780 return true;
781 return false;
782}
783
5ec4e49f
JK		 784static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
785{
786 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
787 return true;
788 return false;
789}
ac5d156c 790
87d6f890
JK		 791/*
792 * It is very important to gather dirty pages and write at once, so that we can
793 * submit a big bio without interfering other data writes.
794 * By default, 512 pages for directory data,
727ebb09
KM		 795 * 512 pages (2MB) * 8 for nodes, and
796 * 256 pages * 8 for meta are set.
87d6f890
JK		 797 */
798static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
799{
a88a341a 800 if (sbi->sb->s_bdi->wb.dirty_exceeded)
510184c8
JK		 801 return 0;
802
a1257023
JK		 803 if (type == DATA)
804 return sbi->blocks_per_seg;
805 else if (type == NODE)
2c237eba 806 return 8 * sbi->blocks_per_seg;
87d6f890 807 else if (type == META)
664ba972 808 return 8 * BIO_MAX_PAGES;
87d6f890
JK		 809 else
810 return 0;
811}
50c8cdb3
JK		 812
813/*
814 * When writing pages, it'd better align nr_to_write for segment size.
815 */
816static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
817 struct writeback_control *wbc)
818{
819 long nr_to_write, desired;
820
821 if (wbc->sync_mode != WB_SYNC_NONE)
822 return 0;
823
824 nr_to_write = wbc->nr_to_write;
664ba972 825 desired = BIO_MAX_PAGES;
28ea6162 826 if (type == NODE)
664ba972 827 desired <<= 1;
50c8cdb3
JK		 828
829 wbc->nr_to_write = desired;
830 return desired - nr_to_write;
831}
01983c71
JK		 832
833static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
834{
835 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
836 bool wakeup = false;
837 int i;
838
839 if (force)
840 goto wake_up;
841
842 mutex_lock(&dcc->cmd_lock);
78997b56
CY		 843 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
844 if (i + 1 < dcc->discard_granularity)
845 break;
01983c71
JK		 846 if (!list_empty(&dcc->pend_list[i])) {
847 wakeup = true;
848 break;
849 }
850 }
851 mutex_unlock(&dcc->cmd_lock);
852 if (!wakeup)
853 return;
854wake_up:
855 dcc->discard_wake = 1;
856 wake_up_interruptible_all(&dcc->discard_wait_queue);
857}
Linux 6.x block layer and io_uring tree(s)