1 /* SPDX-License-Identifier: GPL-2.0 */
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
8 #include <linux/blkdev.h>
9 #include <linux/backing-dev.h>
12 #define NULL_SEGNO ((unsigned int)(~0))
13 #define NULL_SECNO ((unsigned int)(~0))
15 #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
16 #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
18 #define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
19 #define F2FS_MIN_META_SEGMENTS 8 /* SB + 2 (CP + SIT + NAT) + SSA */
21 /* L: Logical segment # in volume, R: Relative segment # in main area */
22 #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
23 #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
25 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
26 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
27 #define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA))
29 static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
30 unsigned short seg_type)
32 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
35 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
36 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
37 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
39 #define IS_CURSEG(sbi, seg) \
40 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
41 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
42 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
43 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
44 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
45 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \
46 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \
47 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
49 #define IS_CURSEC(sbi, secno) \
50 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
51 (sbi)->segs_per_sec) || \
52 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
53 (sbi)->segs_per_sec) || \
54 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
55 (sbi)->segs_per_sec) || \
56 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
57 (sbi)->segs_per_sec) || \
58 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
59 (sbi)->segs_per_sec) || \
60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
61 (sbi)->segs_per_sec) || \
62 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
63 (sbi)->segs_per_sec) || \
64 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
67 #define MAIN_BLKADDR(sbi) \
68 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
69 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
70 #define SEG0_BLKADDR(sbi) \
71 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
72 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
74 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
75 #define MAIN_SECS(sbi) ((sbi)->total_sections)
77 #define TOTAL_SEGS(sbi) \
78 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
79 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
80 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
82 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
83 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
84 (sbi)->log_blocks_per_seg))
86 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
87 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
89 #define NEXT_FREE_BLKADDR(sbi, curseg) \
90 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
92 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
93 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
94 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
95 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
96 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
98 #define GET_SEGNO(sbi, blk_addr) \
99 ((!__is_valid_data_blkaddr(blk_addr)) ? \
100 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
101 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
102 #define BLKS_PER_SEC(sbi) \
103 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
104 #define CAP_BLKS_PER_SEC(sbi) \
105 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg - \
106 (sbi)->unusable_blocks_per_sec)
107 #define CAP_SEGS_PER_SEC(sbi) \
108 ((sbi)->segs_per_sec - ((sbi)->unusable_blocks_per_sec >>\
109 (sbi)->log_blocks_per_seg))
110 #define GET_SEC_FROM_SEG(sbi, segno) \
111 (((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
112 #define GET_SEG_FROM_SEC(sbi, secno) \
113 ((secno) * (sbi)->segs_per_sec)
114 #define GET_ZONE_FROM_SEC(sbi, secno) \
115 (((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
116 #define GET_ZONE_FROM_SEG(sbi, segno) \
117 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
119 #define GET_SUM_BLOCK(sbi, segno) \
120 ((sbi)->sm_info->ssa_blkaddr + (segno))
122 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
123 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
125 #define SIT_ENTRY_OFFSET(sit_i, segno) \
126 ((segno) % (sit_i)->sents_per_block)
127 #define SIT_BLOCK_OFFSET(segno) \
128 ((segno) / SIT_ENTRY_PER_BLOCK)
129 #define START_SEGNO(segno) \
130 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
131 #define SIT_BLK_CNT(sbi) \
132 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
133 #define f2fs_bitmap_size(nr) \
134 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
136 #define SECTOR_FROM_BLOCK(blk_addr) \
137 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
138 #define SECTOR_TO_BLOCK(sectors) \
139 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
142 * indicate a block allocation direction: RIGHT and LEFT.
143 * RIGHT means allocating new sections towards the end of volume.
144 * LEFT means the opposite direction.
152 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
153 * LFS writes data sequentially with cleaning operations.
154 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
155 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
156 * fragmented segment which has similar aging degree.
165 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
166 * GC_CB is based on cost-benefit algorithm.
167 * GC_GREEDY is based on greedy algorithm.
168 * GC_AT is based on age-threshold algorithm.
180 * BG_GC means the background cleaning job.
181 * FG_GC means the on-demand cleaning job.
182 * FORCE_FG_GC means on-demand cleaning job in background.
190 /* for a function parameter to select a victim segment */
191 struct victim_sel_policy {
192 int alloc_mode; /* LFS or SSR */
193 int gc_mode; /* GC_CB or GC_GREEDY */
194 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
195 unsigned int max_search; /*
196 * maximum # of segments/sections
199 unsigned int offset; /* last scanned bitmap offset */
200 unsigned int ofs_unit; /* bitmap search unit */
201 unsigned int min_cost; /* minimum cost */
202 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
203 unsigned int min_segno; /* segment # having min. cost */
204 unsigned long long age; /* mtime of GCed section*/
205 unsigned long long age_threshold;/* age threshold */
209 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
210 unsigned int valid_blocks:10; /* # of valid blocks */
211 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
212 unsigned int padding:6; /* padding */
213 unsigned char *cur_valid_map; /* validity bitmap of blocks */
214 #ifdef CONFIG_F2FS_CHECK_FS
215 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
218 * # of valid blocks and the validity bitmap stored in the last
219 * checkpoint pack. This information is used by the SSR mode.
221 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
222 unsigned char *discard_map;
223 unsigned long long mtime; /* modification time of the segment */
227 unsigned int valid_blocks; /* # of valid blocks in a section */
230 struct segment_allocation {
231 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
234 #define MAX_SKIP_GC_COUNT 16
237 struct list_head list;
239 block_t old_addr; /* for revoking when fail to commit */
243 const struct segment_allocation *s_ops;
245 block_t sit_base_addr; /* start block address of SIT area */
246 block_t sit_blocks; /* # of blocks used by SIT area */
247 block_t written_valid_blocks; /* # of valid blocks in main area */
248 char *bitmap; /* all bitmaps pointer */
249 char *sit_bitmap; /* SIT bitmap pointer */
250 #ifdef CONFIG_F2FS_CHECK_FS
251 char *sit_bitmap_mir; /* SIT bitmap mirror */
253 /* bitmap of segments to be ignored by GC in case of errors */
254 unsigned long *invalid_segmap;
256 unsigned int bitmap_size; /* SIT bitmap size */
258 unsigned long *tmp_map; /* bitmap for temporal use */
259 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
260 unsigned int dirty_sentries; /* # of dirty sentries */
261 unsigned int sents_per_block; /* # of SIT entries per block */
262 struct rw_semaphore sentry_lock; /* to protect SIT cache */
263 struct seg_entry *sentries; /* SIT segment-level cache */
264 struct sec_entry *sec_entries; /* SIT section-level cache */
266 /* for cost-benefit algorithm in cleaning procedure */
267 unsigned long long elapsed_time; /* elapsed time after mount */
268 unsigned long long mounted_time; /* mount time */
269 unsigned long long min_mtime; /* min. modification time */
270 unsigned long long max_mtime; /* max. modification time */
271 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
272 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
274 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
277 struct free_segmap_info {
278 unsigned int start_segno; /* start segment number logically */
279 unsigned int free_segments; /* # of free segments */
280 unsigned int free_sections; /* # of free sections */
281 spinlock_t segmap_lock; /* free segmap lock */
282 unsigned long *free_segmap; /* free segment bitmap */
283 unsigned long *free_secmap; /* free section bitmap */
286 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
288 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
289 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
290 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
291 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
292 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
293 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
294 DIRTY, /* to count # of dirty segments */
295 PRE, /* to count # of entirely obsolete segments */
299 struct dirty_seglist_info {
300 const struct victim_selection *v_ops; /* victim selction operation */
301 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
302 unsigned long *dirty_secmap;
303 struct mutex seglist_lock; /* lock for segment bitmaps */
304 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
305 unsigned long *victim_secmap; /* background GC victims */
308 /* victim selection function for cleaning and SSR */
309 struct victim_selection {
310 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
311 int, int, char, unsigned long long);
314 /* for active log information */
316 struct mutex curseg_mutex; /* lock for consistency */
317 struct f2fs_summary_block *sum_blk; /* cached summary block */
318 struct rw_semaphore journal_rwsem; /* protect journal area */
319 struct f2fs_journal *journal; /* cached journal info */
320 unsigned char alloc_type; /* current allocation type */
321 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
322 unsigned int segno; /* current segment number */
323 unsigned short next_blkoff; /* next block offset to write */
324 unsigned int zone; /* current zone number */
325 unsigned int next_segno; /* preallocated segment */
326 bool inited; /* indicate inmem log is inited */
329 struct sit_entry_set {
330 struct list_head set_list; /* link with all sit sets */
331 unsigned int start_segno; /* start segno of sits in set */
332 unsigned int entry_cnt; /* the # of sit entries in set */
338 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
340 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
343 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
346 struct sit_info *sit_i = SIT_I(sbi);
347 return &sit_i->sentries[segno];
350 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
353 struct sit_info *sit_i = SIT_I(sbi);
354 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
357 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
358 unsigned int segno, bool use_section)
361 * In order to get # of valid blocks in a section instantly from many
362 * segments, f2fs manages two counting structures separately.
364 if (use_section && __is_large_section(sbi))
365 return get_sec_entry(sbi, segno)->valid_blocks;
367 return get_seg_entry(sbi, segno)->valid_blocks;
370 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
371 unsigned int segno, bool use_section)
373 if (use_section && __is_large_section(sbi)) {
374 unsigned int start_segno = START_SEGNO(segno);
375 unsigned int blocks = 0;
378 for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
379 struct seg_entry *se = get_seg_entry(sbi, start_segno);
381 blocks += se->ckpt_valid_blocks;
385 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
388 static inline void seg_info_from_raw_sit(struct seg_entry *se,
389 struct f2fs_sit_entry *rs)
391 se->valid_blocks = GET_SIT_VBLOCKS(rs);
392 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
393 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
394 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
395 #ifdef CONFIG_F2FS_CHECK_FS
396 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
398 se->type = GET_SIT_TYPE(rs);
399 se->mtime = le64_to_cpu(rs->mtime);
402 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
403 struct f2fs_sit_entry *rs)
405 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
407 rs->vblocks = cpu_to_le16(raw_vblocks);
408 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
409 rs->mtime = cpu_to_le64(se->mtime);
412 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
413 struct page *page, unsigned int start)
415 struct f2fs_sit_block *raw_sit;
416 struct seg_entry *se;
417 struct f2fs_sit_entry *rs;
418 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
419 (unsigned long)MAIN_SEGS(sbi));
422 raw_sit = (struct f2fs_sit_block *)page_address(page);
423 memset(raw_sit, 0, PAGE_SIZE);
424 for (i = 0; i < end - start; i++) {
425 rs = &raw_sit->entries[i];
426 se = get_seg_entry(sbi, start + i);
427 __seg_info_to_raw_sit(se, rs);
431 static inline void seg_info_to_raw_sit(struct seg_entry *se,
432 struct f2fs_sit_entry *rs)
434 __seg_info_to_raw_sit(se, rs);
436 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
437 se->ckpt_valid_blocks = se->valid_blocks;
440 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
441 unsigned int max, unsigned int segno)
444 spin_lock(&free_i->segmap_lock);
445 ret = find_next_bit(free_i->free_segmap, max, segno);
446 spin_unlock(&free_i->segmap_lock);
450 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
452 struct free_segmap_info *free_i = FREE_I(sbi);
453 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
454 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
456 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
458 spin_lock(&free_i->segmap_lock);
459 clear_bit(segno, free_i->free_segmap);
460 free_i->free_segments++;
462 next = find_next_bit(free_i->free_segmap,
463 start_segno + sbi->segs_per_sec, start_segno);
464 if (next >= start_segno + usable_segs) {
465 clear_bit(secno, free_i->free_secmap);
466 free_i->free_sections++;
468 spin_unlock(&free_i->segmap_lock);
471 static inline void __set_inuse(struct f2fs_sb_info *sbi,
474 struct free_segmap_info *free_i = FREE_I(sbi);
475 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
477 set_bit(segno, free_i->free_segmap);
478 free_i->free_segments--;
479 if (!test_and_set_bit(secno, free_i->free_secmap))
480 free_i->free_sections--;
483 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
484 unsigned int segno, bool inmem)
486 struct free_segmap_info *free_i = FREE_I(sbi);
487 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
488 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
490 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
492 spin_lock(&free_i->segmap_lock);
493 if (test_and_clear_bit(segno, free_i->free_segmap)) {
494 free_i->free_segments++;
496 if (!inmem && IS_CURSEC(sbi, secno))
498 next = find_next_bit(free_i->free_segmap,
499 start_segno + sbi->segs_per_sec, start_segno);
500 if (next >= start_segno + usable_segs) {
501 if (test_and_clear_bit(secno, free_i->free_secmap))
502 free_i->free_sections++;
506 spin_unlock(&free_i->segmap_lock);
509 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
512 struct free_segmap_info *free_i = FREE_I(sbi);
513 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
515 spin_lock(&free_i->segmap_lock);
516 if (!test_and_set_bit(segno, free_i->free_segmap)) {
517 free_i->free_segments--;
518 if (!test_and_set_bit(secno, free_i->free_secmap))
519 free_i->free_sections--;
521 spin_unlock(&free_i->segmap_lock);
524 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
527 struct sit_info *sit_i = SIT_I(sbi);
529 #ifdef CONFIG_F2FS_CHECK_FS
530 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
534 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
537 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
539 return SIT_I(sbi)->written_valid_blocks;
542 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
544 return FREE_I(sbi)->free_segments;
547 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
549 return SM_I(sbi)->reserved_segments +
550 SM_I(sbi)->additional_reserved_segments;
553 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
555 return FREE_I(sbi)->free_sections;
558 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
560 return DIRTY_I(sbi)->nr_dirty[PRE];
563 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
565 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
566 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
567 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
568 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
569 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
570 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
573 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
575 return SM_I(sbi)->ovp_segments;
578 static inline int reserved_sections(struct f2fs_sb_info *sbi)
580 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
583 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
584 unsigned int node_blocks, unsigned int dent_blocks)
587 unsigned int segno, left_blocks;
590 /* check current node segment */
591 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
592 segno = CURSEG_I(sbi, i)->segno;
593 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
594 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
596 if (node_blocks > left_blocks)
600 /* check current data segment */
601 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
602 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
603 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
604 if (dent_blocks > left_blocks)
609 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
610 int freed, int needed)
612 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
613 get_pages(sbi, F2FS_DIRTY_DENTS) +
614 get_pages(sbi, F2FS_DIRTY_IMETA);
615 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
616 unsigned int node_secs = total_node_blocks / BLKS_PER_SEC(sbi);
617 unsigned int dent_secs = total_dent_blocks / BLKS_PER_SEC(sbi);
618 unsigned int node_blocks = total_node_blocks % BLKS_PER_SEC(sbi);
619 unsigned int dent_blocks = total_dent_blocks % BLKS_PER_SEC(sbi);
620 unsigned int free, need_lower, need_upper;
622 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
625 free = free_sections(sbi) + freed;
626 need_lower = node_secs + dent_secs + reserved_sections(sbi) + needed;
627 need_upper = need_lower + (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
629 if (free > need_upper)
631 else if (free <= need_lower)
633 return !has_curseg_enough_space(sbi, node_blocks, dent_blocks);
636 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
638 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
640 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
645 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
647 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
650 static inline int utilization(struct f2fs_sb_info *sbi)
652 return div_u64((u64)valid_user_blocks(sbi) * 100,
653 sbi->user_block_count);
657 * Sometimes f2fs may be better to drop out-of-place update policy.
658 * And, users can control the policy through sysfs entries.
659 * There are five policies with triggering conditions as follows.
660 * F2FS_IPU_FORCE - all the time,
661 * F2FS_IPU_SSR - if SSR mode is activated,
662 * F2FS_IPU_UTIL - if FS utilization is over threashold,
663 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
665 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
666 * storages. IPU will be triggered only if the # of dirty
667 * pages over min_fsync_blocks. (=default option)
668 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
669 * F2FS_IPU_NOCACHE - disable IPU bio cache.
670 * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
672 #define DEF_MIN_IPU_UTIL 70
673 #define DEF_MIN_FSYNC_BLOCKS 8
674 #define DEF_MIN_HOT_BLOCKS 16
676 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
688 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
691 struct curseg_info *curseg = CURSEG_I(sbi, type);
692 return curseg->segno;
695 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
698 struct curseg_info *curseg = CURSEG_I(sbi, type);
699 return curseg->alloc_type;
702 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
704 struct curseg_info *curseg = CURSEG_I(sbi, type);
705 return curseg->next_blkoff;
708 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
710 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
713 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
715 struct f2fs_sb_info *sbi = fio->sbi;
717 if (__is_valid_data_blkaddr(fio->old_blkaddr))
718 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
719 META_GENERIC : DATA_GENERIC);
720 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
721 META_GENERIC : DATA_GENERIC_ENHANCE);
725 * Summary block is always treated as an invalid block
727 static inline int check_block_count(struct f2fs_sb_info *sbi,
728 int segno, struct f2fs_sit_entry *raw_sit)
730 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
731 int valid_blocks = 0;
732 int cur_pos = 0, next_pos;
733 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
735 /* check bitmap with valid block count */
738 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
741 valid_blocks += next_pos - cur_pos;
743 next_pos = find_next_bit_le(&raw_sit->valid_map,
747 is_valid = !is_valid;
748 } while (cur_pos < usable_blks_per_seg);
750 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
751 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
752 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
753 set_sbi_flag(sbi, SBI_NEED_FSCK);
754 return -EFSCORRUPTED;
757 if (usable_blks_per_seg < sbi->blocks_per_seg)
758 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
760 usable_blks_per_seg) != sbi->blocks_per_seg);
762 /* check segment usage, and check boundary of a given segment number */
763 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
764 || segno > TOTAL_SEGS(sbi) - 1)) {
765 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
766 GET_SIT_VBLOCKS(raw_sit), segno);
767 set_sbi_flag(sbi, SBI_NEED_FSCK);
768 return -EFSCORRUPTED;
773 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
776 struct sit_info *sit_i = SIT_I(sbi);
777 unsigned int offset = SIT_BLOCK_OFFSET(start);
778 block_t blk_addr = sit_i->sit_base_addr + offset;
780 check_seg_range(sbi, start);
782 #ifdef CONFIG_F2FS_CHECK_FS
783 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
784 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
788 /* calculate sit block address */
789 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
790 blk_addr += sit_i->sit_blocks;
795 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
798 struct sit_info *sit_i = SIT_I(sbi);
799 block_addr -= sit_i->sit_base_addr;
800 if (block_addr < sit_i->sit_blocks)
801 block_addr += sit_i->sit_blocks;
803 block_addr -= sit_i->sit_blocks;
805 return block_addr + sit_i->sit_base_addr;
808 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
810 unsigned int block_off = SIT_BLOCK_OFFSET(start);
812 f2fs_change_bit(block_off, sit_i->sit_bitmap);
813 #ifdef CONFIG_F2FS_CHECK_FS
814 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
818 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
821 struct sit_info *sit_i = SIT_I(sbi);
822 time64_t diff, now = ktime_get_boottime_seconds();
824 if (now >= sit_i->mounted_time)
825 return sit_i->elapsed_time + now - sit_i->mounted_time;
827 /* system time is set to the past */
829 diff = sit_i->mounted_time - now;
830 if (sit_i->elapsed_time >= diff)
831 return sit_i->elapsed_time - diff;
834 return sit_i->elapsed_time;
837 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
838 unsigned int ofs_in_node, unsigned char version)
840 sum->nid = cpu_to_le32(nid);
841 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
842 sum->version = version;
845 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
847 return __start_cp_addr(sbi) +
848 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
851 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
853 return __start_cp_addr(sbi) +
854 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
858 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
860 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
866 * It is very important to gather dirty pages and write at once, so that we can
867 * submit a big bio without interfering other data writes.
868 * By default, 512 pages for directory data,
869 * 512 pages (2MB) * 8 for nodes, and
870 * 256 pages * 8 for meta are set.
872 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
874 if (sbi->sb->s_bdi->wb.dirty_exceeded)
878 return sbi->blocks_per_seg;
879 else if (type == NODE)
880 return 8 * sbi->blocks_per_seg;
881 else if (type == META)
882 return 8 * BIO_MAX_PAGES;
888 * When writing pages, it'd better align nr_to_write for segment size.
890 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
891 struct writeback_control *wbc)
893 long nr_to_write, desired;
895 if (wbc->sync_mode != WB_SYNC_NONE)
898 nr_to_write = wbc->nr_to_write;
899 desired = BIO_MAX_PAGES;
903 wbc->nr_to_write = desired;
904 return desired - nr_to_write;
907 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
909 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
916 mutex_lock(&dcc->cmd_lock);
917 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
918 if (i + 1 < dcc->discard_granularity)
920 if (!list_empty(&dcc->pend_list[i])) {
925 mutex_unlock(&dcc->cmd_lock);
926 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
929 dcc->discard_wake = 1;
930 wake_up_interruptible_all(&dcc->discard_wait_queue);