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)
28 static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
29 unsigned short seg_type)
31 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
34 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
35 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
36 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
38 #define IS_CURSEG(sbi, seg) \
39 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
40 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
41 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
42 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
43 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
44 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \
45 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \
46 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
48 #define IS_CURSEC(sbi, secno) \
49 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
50 (sbi)->segs_per_sec) || \
51 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
52 (sbi)->segs_per_sec) || \
53 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
54 (sbi)->segs_per_sec) || \
55 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
56 (sbi)->segs_per_sec) || \
57 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
58 (sbi)->segs_per_sec) || \
59 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
60 (sbi)->segs_per_sec) || \
61 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
62 (sbi)->segs_per_sec) || \
63 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
66 #define MAIN_BLKADDR(sbi) \
67 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
68 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
69 #define SEG0_BLKADDR(sbi) \
70 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
71 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
73 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
74 #define MAIN_SECS(sbi) ((sbi)->total_sections)
76 #define TOTAL_SEGS(sbi) \
77 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
78 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
79 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
81 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
82 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
83 (sbi)->log_blocks_per_seg))
85 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
86 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
88 #define NEXT_FREE_BLKADDR(sbi, curseg) \
89 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
91 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
92 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
93 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
94 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
95 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
97 #define GET_SEGNO(sbi, blk_addr) \
98 ((!__is_valid_data_blkaddr(blk_addr)) ? \
99 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
100 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
101 #define BLKS_PER_SEC(sbi) \
102 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
103 #define GET_SEC_FROM_SEG(sbi, segno) \
104 (((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
105 #define GET_SEG_FROM_SEC(sbi, secno) \
106 ((secno) * (sbi)->segs_per_sec)
107 #define GET_ZONE_FROM_SEC(sbi, secno) \
108 (((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
109 #define GET_ZONE_FROM_SEG(sbi, segno) \
110 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
112 #define GET_SUM_BLOCK(sbi, segno) \
113 ((sbi)->sm_info->ssa_blkaddr + (segno))
115 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
116 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
118 #define SIT_ENTRY_OFFSET(sit_i, segno) \
119 ((segno) % (sit_i)->sents_per_block)
120 #define SIT_BLOCK_OFFSET(segno) \
121 ((segno) / SIT_ENTRY_PER_BLOCK)
122 #define START_SEGNO(segno) \
123 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
124 #define SIT_BLK_CNT(sbi) \
125 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
126 #define f2fs_bitmap_size(nr) \
127 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
129 #define SECTOR_FROM_BLOCK(blk_addr) \
130 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
131 #define SECTOR_TO_BLOCK(sectors) \
132 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
135 * indicate a block allocation direction: RIGHT and LEFT.
136 * RIGHT means allocating new sections towards the end of volume.
137 * LEFT means the opposite direction.
145 * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
146 * LFS writes data sequentially with cleaning operations.
147 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
148 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
149 * fragmented segment which has similar aging degree.
158 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
159 * GC_CB is based on cost-benefit algorithm.
160 * GC_GREEDY is based on greedy algorithm.
161 * GC_AT is based on age-threshold algorithm.
173 * BG_GC means the background cleaning job.
174 * FG_GC means the on-demand cleaning job.
181 /* for a function parameter to select a victim segment */
182 struct victim_sel_policy {
183 int alloc_mode; /* LFS or SSR */
184 int gc_mode; /* GC_CB or GC_GREEDY */
185 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
186 unsigned int max_search; /*
187 * maximum # of segments/sections
190 unsigned int offset; /* last scanned bitmap offset */
191 unsigned int ofs_unit; /* bitmap search unit */
192 unsigned int min_cost; /* minimum cost */
193 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
194 unsigned int min_segno; /* segment # having min. cost */
195 unsigned long long age; /* mtime of GCed section*/
196 unsigned long long age_threshold;/* age threshold */
200 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
201 unsigned int valid_blocks:10; /* # of valid blocks */
202 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
203 unsigned int padding:6; /* padding */
204 unsigned char *cur_valid_map; /* validity bitmap of blocks */
205 #ifdef CONFIG_F2FS_CHECK_FS
206 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
209 * # of valid blocks and the validity bitmap stored in the last
210 * checkpoint pack. This information is used by the SSR mode.
212 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
213 unsigned char *discard_map;
214 unsigned long long mtime; /* modification time of the segment */
218 unsigned int valid_blocks; /* # of valid blocks in a section */
221 struct segment_allocation {
222 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
225 #define MAX_SKIP_GC_COUNT 16
228 struct list_head list;
230 block_t old_addr; /* for revoking when fail to commit */
234 const struct segment_allocation *s_ops;
236 block_t sit_base_addr; /* start block address of SIT area */
237 block_t sit_blocks; /* # of blocks used by SIT area */
238 block_t written_valid_blocks; /* # of valid blocks in main area */
239 char *bitmap; /* all bitmaps pointer */
240 char *sit_bitmap; /* SIT bitmap pointer */
241 #ifdef CONFIG_F2FS_CHECK_FS
242 char *sit_bitmap_mir; /* SIT bitmap mirror */
244 /* bitmap of segments to be ignored by GC in case of errors */
245 unsigned long *invalid_segmap;
247 unsigned int bitmap_size; /* SIT bitmap size */
249 unsigned long *tmp_map; /* bitmap for temporal use */
250 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
251 unsigned int dirty_sentries; /* # of dirty sentries */
252 unsigned int sents_per_block; /* # of SIT entries per block */
253 struct rw_semaphore sentry_lock; /* to protect SIT cache */
254 struct seg_entry *sentries; /* SIT segment-level cache */
255 struct sec_entry *sec_entries; /* SIT section-level cache */
257 /* for cost-benefit algorithm in cleaning procedure */
258 unsigned long long elapsed_time; /* elapsed time after mount */
259 unsigned long long mounted_time; /* mount time */
260 unsigned long long min_mtime; /* min. modification time */
261 unsigned long long max_mtime; /* max. modification time */
262 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
263 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
265 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
268 struct free_segmap_info {
269 unsigned int start_segno; /* start segment number logically */
270 unsigned int free_segments; /* # of free segments */
271 unsigned int free_sections; /* # of free sections */
272 spinlock_t segmap_lock; /* free segmap lock */
273 unsigned long *free_segmap; /* free segment bitmap */
274 unsigned long *free_secmap; /* free section bitmap */
277 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
279 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
280 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
281 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
282 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
283 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
284 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
285 DIRTY, /* to count # of dirty segments */
286 PRE, /* to count # of entirely obsolete segments */
290 struct dirty_seglist_info {
291 const struct victim_selection *v_ops; /* victim selction operation */
292 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
293 unsigned long *dirty_secmap;
294 struct mutex seglist_lock; /* lock for segment bitmaps */
295 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
296 unsigned long *victim_secmap; /* background GC victims */
299 /* victim selection function for cleaning and SSR */
300 struct victim_selection {
301 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
302 int, int, char, unsigned long long);
305 /* for active log information */
307 struct mutex curseg_mutex; /* lock for consistency */
308 struct f2fs_summary_block *sum_blk; /* cached summary block */
309 struct rw_semaphore journal_rwsem; /* protect journal area */
310 struct f2fs_journal *journal; /* cached journal info */
311 unsigned char alloc_type; /* current allocation type */
312 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
313 unsigned int segno; /* current segment number */
314 unsigned short next_blkoff; /* next block offset to write */
315 unsigned int zone; /* current zone number */
316 unsigned int next_segno; /* preallocated segment */
317 bool inited; /* indicate inmem log is inited */
320 struct sit_entry_set {
321 struct list_head set_list; /* link with all sit sets */
322 unsigned int start_segno; /* start segno of sits in set */
323 unsigned int entry_cnt; /* the # of sit entries in set */
329 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
331 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
334 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
337 struct sit_info *sit_i = SIT_I(sbi);
338 return &sit_i->sentries[segno];
341 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
344 struct sit_info *sit_i = SIT_I(sbi);
345 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
348 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
349 unsigned int segno, bool use_section)
352 * In order to get # of valid blocks in a section instantly from many
353 * segments, f2fs manages two counting structures separately.
355 if (use_section && __is_large_section(sbi))
356 return get_sec_entry(sbi, segno)->valid_blocks;
358 return get_seg_entry(sbi, segno)->valid_blocks;
361 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
362 unsigned int segno, bool use_section)
364 if (use_section && __is_large_section(sbi)) {
365 unsigned int start_segno = START_SEGNO(segno);
366 unsigned int blocks = 0;
369 for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
370 struct seg_entry *se = get_seg_entry(sbi, start_segno);
372 blocks += se->ckpt_valid_blocks;
376 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
379 static inline void seg_info_from_raw_sit(struct seg_entry *se,
380 struct f2fs_sit_entry *rs)
382 se->valid_blocks = GET_SIT_VBLOCKS(rs);
383 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
384 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
385 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
386 #ifdef CONFIG_F2FS_CHECK_FS
387 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
389 se->type = GET_SIT_TYPE(rs);
390 se->mtime = le64_to_cpu(rs->mtime);
393 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
394 struct f2fs_sit_entry *rs)
396 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
398 rs->vblocks = cpu_to_le16(raw_vblocks);
399 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
400 rs->mtime = cpu_to_le64(se->mtime);
403 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
404 struct page *page, unsigned int start)
406 struct f2fs_sit_block *raw_sit;
407 struct seg_entry *se;
408 struct f2fs_sit_entry *rs;
409 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
410 (unsigned long)MAIN_SEGS(sbi));
413 raw_sit = (struct f2fs_sit_block *)page_address(page);
414 memset(raw_sit, 0, PAGE_SIZE);
415 for (i = 0; i < end - start; i++) {
416 rs = &raw_sit->entries[i];
417 se = get_seg_entry(sbi, start + i);
418 __seg_info_to_raw_sit(se, rs);
422 static inline void seg_info_to_raw_sit(struct seg_entry *se,
423 struct f2fs_sit_entry *rs)
425 __seg_info_to_raw_sit(se, rs);
427 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
428 se->ckpt_valid_blocks = se->valid_blocks;
431 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
432 unsigned int max, unsigned int segno)
435 spin_lock(&free_i->segmap_lock);
436 ret = find_next_bit(free_i->free_segmap, max, segno);
437 spin_unlock(&free_i->segmap_lock);
441 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
443 struct free_segmap_info *free_i = FREE_I(sbi);
444 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
445 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
447 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
449 spin_lock(&free_i->segmap_lock);
450 clear_bit(segno, free_i->free_segmap);
451 free_i->free_segments++;
453 next = find_next_bit(free_i->free_segmap,
454 start_segno + sbi->segs_per_sec, start_segno);
455 if (next >= start_segno + usable_segs) {
456 clear_bit(secno, free_i->free_secmap);
457 free_i->free_sections++;
459 spin_unlock(&free_i->segmap_lock);
462 static inline void __set_inuse(struct f2fs_sb_info *sbi,
465 struct free_segmap_info *free_i = FREE_I(sbi);
466 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
468 set_bit(segno, free_i->free_segmap);
469 free_i->free_segments--;
470 if (!test_and_set_bit(secno, free_i->free_secmap))
471 free_i->free_sections--;
474 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
475 unsigned int segno, bool inmem)
477 struct free_segmap_info *free_i = FREE_I(sbi);
478 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
479 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
481 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
483 spin_lock(&free_i->segmap_lock);
484 if (test_and_clear_bit(segno, free_i->free_segmap)) {
485 free_i->free_segments++;
487 if (!inmem && IS_CURSEC(sbi, secno))
489 next = find_next_bit(free_i->free_segmap,
490 start_segno + sbi->segs_per_sec, start_segno);
491 if (next >= start_segno + usable_segs) {
492 if (test_and_clear_bit(secno, free_i->free_secmap))
493 free_i->free_sections++;
497 spin_unlock(&free_i->segmap_lock);
500 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
503 struct free_segmap_info *free_i = FREE_I(sbi);
504 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
506 spin_lock(&free_i->segmap_lock);
507 if (!test_and_set_bit(segno, free_i->free_segmap)) {
508 free_i->free_segments--;
509 if (!test_and_set_bit(secno, free_i->free_secmap))
510 free_i->free_sections--;
512 spin_unlock(&free_i->segmap_lock);
515 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
518 struct sit_info *sit_i = SIT_I(sbi);
520 #ifdef CONFIG_F2FS_CHECK_FS
521 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
525 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
528 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
530 return SIT_I(sbi)->written_valid_blocks;
533 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
535 return FREE_I(sbi)->free_segments;
538 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
540 return SM_I(sbi)->reserved_segments;
543 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
545 return FREE_I(sbi)->free_sections;
548 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
550 return DIRTY_I(sbi)->nr_dirty[PRE];
553 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
555 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
556 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
557 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
558 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
559 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
560 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
563 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
565 return SM_I(sbi)->ovp_segments;
568 static inline int reserved_sections(struct f2fs_sb_info *sbi)
570 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
573 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
575 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
576 get_pages(sbi, F2FS_DIRTY_DENTS);
577 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
578 unsigned int segno, left_blocks;
581 /* check current node segment */
582 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
583 segno = CURSEG_I(sbi, i)->segno;
584 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
585 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
587 if (node_blocks > left_blocks)
591 /* check current data segment */
592 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
593 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
594 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
595 if (dent_blocks > left_blocks)
600 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
601 int freed, int needed)
603 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
604 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
605 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
607 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
610 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
611 has_curseg_enough_space(sbi))
613 return (free_sections(sbi) + freed) <=
614 (node_secs + 2 * dent_secs + imeta_secs +
615 reserved_sections(sbi) + needed);
618 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
620 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
622 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
627 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
629 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
632 static inline int utilization(struct f2fs_sb_info *sbi)
634 return div_u64((u64)valid_user_blocks(sbi) * 100,
635 sbi->user_block_count);
639 * Sometimes f2fs may be better to drop out-of-place update policy.
640 * And, users can control the policy through sysfs entries.
641 * There are five policies with triggering conditions as follows.
642 * F2FS_IPU_FORCE - all the time,
643 * F2FS_IPU_SSR - if SSR mode is activated,
644 * F2FS_IPU_UTIL - if FS utilization is over threashold,
645 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
647 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
648 * storages. IPU will be triggered only if the # of dirty
649 * pages over min_fsync_blocks. (=default option)
650 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
651 * F2FS_IPU_NOCACHE - disable IPU bio cache.
652 * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
654 #define DEF_MIN_IPU_UTIL 70
655 #define DEF_MIN_FSYNC_BLOCKS 8
656 #define DEF_MIN_HOT_BLOCKS 16
658 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
670 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
673 struct curseg_info *curseg = CURSEG_I(sbi, type);
674 return curseg->segno;
677 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
680 struct curseg_info *curseg = CURSEG_I(sbi, type);
681 return curseg->alloc_type;
684 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
686 struct curseg_info *curseg = CURSEG_I(sbi, type);
687 return curseg->next_blkoff;
690 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
692 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
695 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
697 struct f2fs_sb_info *sbi = fio->sbi;
699 if (__is_valid_data_blkaddr(fio->old_blkaddr))
700 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
701 META_GENERIC : DATA_GENERIC);
702 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
703 META_GENERIC : DATA_GENERIC_ENHANCE);
707 * Summary block is always treated as an invalid block
709 static inline int check_block_count(struct f2fs_sb_info *sbi,
710 int segno, struct f2fs_sit_entry *raw_sit)
712 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
713 int valid_blocks = 0;
714 int cur_pos = 0, next_pos;
715 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
717 /* check bitmap with valid block count */
720 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
723 valid_blocks += next_pos - cur_pos;
725 next_pos = find_next_bit_le(&raw_sit->valid_map,
729 is_valid = !is_valid;
730 } while (cur_pos < usable_blks_per_seg);
732 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
733 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
734 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
735 set_sbi_flag(sbi, SBI_NEED_FSCK);
736 return -EFSCORRUPTED;
739 if (usable_blks_per_seg < sbi->blocks_per_seg)
740 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
742 usable_blks_per_seg) != sbi->blocks_per_seg);
744 /* check segment usage, and check boundary of a given segment number */
745 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
746 || segno > TOTAL_SEGS(sbi) - 1)) {
747 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
748 GET_SIT_VBLOCKS(raw_sit), segno);
749 set_sbi_flag(sbi, SBI_NEED_FSCK);
750 return -EFSCORRUPTED;
755 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
758 struct sit_info *sit_i = SIT_I(sbi);
759 unsigned int offset = SIT_BLOCK_OFFSET(start);
760 block_t blk_addr = sit_i->sit_base_addr + offset;
762 check_seg_range(sbi, start);
764 #ifdef CONFIG_F2FS_CHECK_FS
765 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
766 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
770 /* calculate sit block address */
771 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
772 blk_addr += sit_i->sit_blocks;
777 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
780 struct sit_info *sit_i = SIT_I(sbi);
781 block_addr -= sit_i->sit_base_addr;
782 if (block_addr < sit_i->sit_blocks)
783 block_addr += sit_i->sit_blocks;
785 block_addr -= sit_i->sit_blocks;
787 return block_addr + sit_i->sit_base_addr;
790 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
792 unsigned int block_off = SIT_BLOCK_OFFSET(start);
794 f2fs_change_bit(block_off, sit_i->sit_bitmap);
795 #ifdef CONFIG_F2FS_CHECK_FS
796 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
800 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
803 struct sit_info *sit_i = SIT_I(sbi);
804 time64_t diff, now = ktime_get_boottime_seconds();
806 if (now >= sit_i->mounted_time)
807 return sit_i->elapsed_time + now - sit_i->mounted_time;
809 /* system time is set to the past */
811 diff = sit_i->mounted_time - now;
812 if (sit_i->elapsed_time >= diff)
813 return sit_i->elapsed_time - diff;
816 return sit_i->elapsed_time;
819 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
820 unsigned int ofs_in_node, unsigned char version)
822 sum->nid = cpu_to_le32(nid);
823 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
824 sum->version = version;
827 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
829 return __start_cp_addr(sbi) +
830 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
833 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
835 return __start_cp_addr(sbi) +
836 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
840 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
842 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
848 * It is very important to gather dirty pages and write at once, so that we can
849 * submit a big bio without interfering other data writes.
850 * By default, 512 pages for directory data,
851 * 512 pages (2MB) * 8 for nodes, and
852 * 256 pages * 8 for meta are set.
854 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
856 if (sbi->sb->s_bdi->wb.dirty_exceeded)
860 return sbi->blocks_per_seg;
861 else if (type == NODE)
862 return 8 * sbi->blocks_per_seg;
863 else if (type == META)
864 return 8 * BIO_MAX_VECS;
870 * When writing pages, it'd better align nr_to_write for segment size.
872 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
873 struct writeback_control *wbc)
875 long nr_to_write, desired;
877 if (wbc->sync_mode != WB_SYNC_NONE)
880 nr_to_write = wbc->nr_to_write;
881 desired = BIO_MAX_VECS;
885 wbc->nr_to_write = desired;
886 return desired - nr_to_write;
889 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
891 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
898 mutex_lock(&dcc->cmd_lock);
899 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
900 if (i + 1 < dcc->discard_granularity)
902 if (!list_empty(&dcc->pend_list[i])) {
907 mutex_unlock(&dcc->cmd_lock);
908 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
911 dcc->discard_wake = 1;
912 wake_up_interruptible_all(&dcc->discard_wait_queue);