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 three 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 three 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.
188 /* for a function parameter to select a victim segment */
189 struct victim_sel_policy {
190 int alloc_mode; /* LFS or SSR */
191 int gc_mode; /* GC_CB or GC_GREEDY */
192 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
193 unsigned int max_search; /*
194 * maximum # of segments/sections
197 unsigned int offset; /* last scanned bitmap offset */
198 unsigned int ofs_unit; /* bitmap search unit */
199 unsigned int min_cost; /* minimum cost */
200 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
201 unsigned int min_segno; /* segment # having min. cost */
202 unsigned long long age; /* mtime of GCed section*/
203 unsigned long long age_threshold;/* age threshold */
207 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
208 unsigned int valid_blocks:10; /* # of valid blocks */
209 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
210 unsigned int padding:6; /* padding */
211 unsigned char *cur_valid_map; /* validity bitmap of blocks */
212 #ifdef CONFIG_F2FS_CHECK_FS
213 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
216 * # of valid blocks and the validity bitmap stored in the last
217 * checkpoint pack. This information is used by the SSR mode.
219 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
220 unsigned char *discard_map;
221 unsigned long long mtime; /* modification time of the segment */
225 unsigned int valid_blocks; /* # of valid blocks in a section */
228 struct segment_allocation {
229 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
232 #define MAX_SKIP_GC_COUNT 16
234 struct revoke_entry {
235 struct list_head list;
236 block_t old_addr; /* for revoking when fail to commit */
241 const struct segment_allocation *s_ops;
243 block_t sit_base_addr; /* start block address of SIT area */
244 block_t sit_blocks; /* # of blocks used by SIT area */
245 block_t written_valid_blocks; /* # of valid blocks in main area */
246 char *bitmap; /* all bitmaps pointer */
247 char *sit_bitmap; /* SIT bitmap pointer */
248 #ifdef CONFIG_F2FS_CHECK_FS
249 char *sit_bitmap_mir; /* SIT bitmap mirror */
251 /* bitmap of segments to be ignored by GC in case of errors */
252 unsigned long *invalid_segmap;
254 unsigned int bitmap_size; /* SIT bitmap size */
256 unsigned long *tmp_map; /* bitmap for temporal use */
257 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
258 unsigned int dirty_sentries; /* # of dirty sentries */
259 unsigned int sents_per_block; /* # of SIT entries per block */
260 struct rw_semaphore sentry_lock; /* to protect SIT cache */
261 struct seg_entry *sentries; /* SIT segment-level cache */
262 struct sec_entry *sec_entries; /* SIT section-level cache */
264 /* for cost-benefit algorithm in cleaning procedure */
265 unsigned long long elapsed_time; /* elapsed time after mount */
266 unsigned long long mounted_time; /* mount time */
267 unsigned long long min_mtime; /* min. modification time */
268 unsigned long long max_mtime; /* max. modification time */
269 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
270 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
272 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
275 struct free_segmap_info {
276 unsigned int start_segno; /* start segment number logically */
277 unsigned int free_segments; /* # of free segments */
278 unsigned int free_sections; /* # of free sections */
279 spinlock_t segmap_lock; /* free segmap lock */
280 unsigned long *free_segmap; /* free segment bitmap */
281 unsigned long *free_secmap; /* free section bitmap */
284 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
286 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
287 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
288 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
289 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
290 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
291 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
292 DIRTY, /* to count # of dirty segments */
293 PRE, /* to count # of entirely obsolete segments */
297 struct dirty_seglist_info {
298 const struct victim_selection *v_ops; /* victim selction operation */
299 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
300 unsigned long *dirty_secmap;
301 struct mutex seglist_lock; /* lock for segment bitmaps */
302 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
303 unsigned long *victim_secmap; /* background GC victims */
304 unsigned long *pinned_secmap; /* pinned victims from foreground GC */
305 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */
306 bool enable_pin_section; /* enable pinning section */
309 /* victim selection function for cleaning and SSR */
310 struct victim_selection {
311 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
312 int, int, char, unsigned long long);
315 /* for active log information */
317 struct mutex curseg_mutex; /* lock for consistency */
318 struct f2fs_summary_block *sum_blk; /* cached summary block */
319 struct rw_semaphore journal_rwsem; /* protect journal area */
320 struct f2fs_journal *journal; /* cached journal info */
321 unsigned char alloc_type; /* current allocation type */
322 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
323 unsigned int segno; /* current segment number */
324 unsigned short next_blkoff; /* next block offset to write */
325 unsigned int zone; /* current zone number */
326 unsigned int next_segno; /* preallocated segment */
327 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */
328 bool inited; /* indicate inmem log is inited */
331 struct sit_entry_set {
332 struct list_head set_list; /* link with all sit sets */
333 unsigned int start_segno; /* start segno of sits in set */
334 unsigned int entry_cnt; /* the # of sit entries in set */
340 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
342 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
345 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
348 struct sit_info *sit_i = SIT_I(sbi);
349 return &sit_i->sentries[segno];
352 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
355 struct sit_info *sit_i = SIT_I(sbi);
356 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
359 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
360 unsigned int segno, bool use_section)
363 * In order to get # of valid blocks in a section instantly from many
364 * segments, f2fs manages two counting structures separately.
366 if (use_section && __is_large_section(sbi))
367 return get_sec_entry(sbi, segno)->valid_blocks;
369 return get_seg_entry(sbi, segno)->valid_blocks;
372 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
373 unsigned int segno, bool use_section)
375 if (use_section && __is_large_section(sbi)) {
376 unsigned int start_segno = START_SEGNO(segno);
377 unsigned int blocks = 0;
380 for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
381 struct seg_entry *se = get_seg_entry(sbi, start_segno);
383 blocks += se->ckpt_valid_blocks;
387 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
390 static inline void seg_info_from_raw_sit(struct seg_entry *se,
391 struct f2fs_sit_entry *rs)
393 se->valid_blocks = GET_SIT_VBLOCKS(rs);
394 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
395 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
396 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
397 #ifdef CONFIG_F2FS_CHECK_FS
398 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
400 se->type = GET_SIT_TYPE(rs);
401 se->mtime = le64_to_cpu(rs->mtime);
404 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
405 struct f2fs_sit_entry *rs)
407 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
409 rs->vblocks = cpu_to_le16(raw_vblocks);
410 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
411 rs->mtime = cpu_to_le64(se->mtime);
414 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
415 struct page *page, unsigned int start)
417 struct f2fs_sit_block *raw_sit;
418 struct seg_entry *se;
419 struct f2fs_sit_entry *rs;
420 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
421 (unsigned long)MAIN_SEGS(sbi));
424 raw_sit = (struct f2fs_sit_block *)page_address(page);
425 memset(raw_sit, 0, PAGE_SIZE);
426 for (i = 0; i < end - start; i++) {
427 rs = &raw_sit->entries[i];
428 se = get_seg_entry(sbi, start + i);
429 __seg_info_to_raw_sit(se, rs);
433 static inline void seg_info_to_raw_sit(struct seg_entry *se,
434 struct f2fs_sit_entry *rs)
436 __seg_info_to_raw_sit(se, rs);
438 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
439 se->ckpt_valid_blocks = se->valid_blocks;
442 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
443 unsigned int max, unsigned int segno)
446 spin_lock(&free_i->segmap_lock);
447 ret = find_next_bit(free_i->free_segmap, max, segno);
448 spin_unlock(&free_i->segmap_lock);
452 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
454 struct free_segmap_info *free_i = FREE_I(sbi);
455 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
456 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
458 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
460 spin_lock(&free_i->segmap_lock);
461 clear_bit(segno, free_i->free_segmap);
462 free_i->free_segments++;
464 next = find_next_bit(free_i->free_segmap,
465 start_segno + sbi->segs_per_sec, start_segno);
466 if (next >= start_segno + usable_segs) {
467 clear_bit(secno, free_i->free_secmap);
468 free_i->free_sections++;
470 spin_unlock(&free_i->segmap_lock);
473 static inline void __set_inuse(struct f2fs_sb_info *sbi,
476 struct free_segmap_info *free_i = FREE_I(sbi);
477 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
479 set_bit(segno, free_i->free_segmap);
480 free_i->free_segments--;
481 if (!test_and_set_bit(secno, free_i->free_secmap))
482 free_i->free_sections--;
485 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
486 unsigned int segno, bool inmem)
488 struct free_segmap_info *free_i = FREE_I(sbi);
489 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
490 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
492 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
494 spin_lock(&free_i->segmap_lock);
495 if (test_and_clear_bit(segno, free_i->free_segmap)) {
496 free_i->free_segments++;
498 if (!inmem && IS_CURSEC(sbi, secno))
500 next = find_next_bit(free_i->free_segmap,
501 start_segno + sbi->segs_per_sec, start_segno);
502 if (next >= start_segno + usable_segs) {
503 if (test_and_clear_bit(secno, free_i->free_secmap))
504 free_i->free_sections++;
508 spin_unlock(&free_i->segmap_lock);
511 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
514 struct free_segmap_info *free_i = FREE_I(sbi);
515 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
517 spin_lock(&free_i->segmap_lock);
518 if (!test_and_set_bit(segno, free_i->free_segmap)) {
519 free_i->free_segments--;
520 if (!test_and_set_bit(secno, free_i->free_secmap))
521 free_i->free_sections--;
523 spin_unlock(&free_i->segmap_lock);
526 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
529 struct sit_info *sit_i = SIT_I(sbi);
531 #ifdef CONFIG_F2FS_CHECK_FS
532 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
536 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
539 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
541 return SIT_I(sbi)->written_valid_blocks;
544 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
546 return FREE_I(sbi)->free_segments;
549 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
551 return SM_I(sbi)->reserved_segments +
552 SM_I(sbi)->additional_reserved_segments;
555 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
557 return FREE_I(sbi)->free_sections;
560 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
562 return DIRTY_I(sbi)->nr_dirty[PRE];
565 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
567 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
568 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
569 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
570 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
571 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
572 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
575 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
577 return SM_I(sbi)->ovp_segments;
580 static inline int reserved_sections(struct f2fs_sb_info *sbi)
582 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
585 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
586 unsigned int node_blocks, unsigned int dent_blocks)
589 unsigned int segno, left_blocks;
592 /* check current node segment */
593 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
594 segno = CURSEG_I(sbi, i)->segno;
595 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
596 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
598 if (node_blocks > left_blocks)
602 /* check current data segment */
603 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
604 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
605 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
606 if (dent_blocks > left_blocks)
612 * calculate needed sections for dirty node/dentry
613 * and call has_curseg_enough_space
615 static inline void __get_secs_required(struct f2fs_sb_info *sbi,
616 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p)
618 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
619 get_pages(sbi, F2FS_DIRTY_DENTS) +
620 get_pages(sbi, F2FS_DIRTY_IMETA);
621 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
622 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
623 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
624 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
625 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
628 *lower_p = node_secs + dent_secs;
630 *upper_p = node_secs + dent_secs +
631 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
633 *curseg_p = has_curseg_enough_space(sbi,
634 node_blocks, dent_blocks);
637 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
638 int freed, int needed)
640 unsigned int free_secs, lower_secs, upper_secs;
643 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
646 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space);
648 free_secs = free_sections(sbi) + freed;
649 lower_secs += needed + reserved_sections(sbi);
650 upper_secs += needed + reserved_sections(sbi);
652 if (free_secs > upper_secs)
654 else if (free_secs <= lower_secs)
656 return !curseg_space;
659 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
661 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
663 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
668 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
670 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
673 static inline int utilization(struct f2fs_sb_info *sbi)
675 return div_u64((u64)valid_user_blocks(sbi) * 100,
676 sbi->user_block_count);
680 * Sometimes f2fs may be better to drop out-of-place update policy.
681 * And, users can control the policy through sysfs entries.
682 * There are five policies with triggering conditions as follows.
683 * F2FS_IPU_FORCE - all the time,
684 * F2FS_IPU_SSR - if SSR mode is activated,
685 * F2FS_IPU_UTIL - if FS utilization is over threashold,
686 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
688 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
689 * storages. IPU will be triggered only if the # of dirty
690 * pages over min_fsync_blocks. (=default option)
691 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
692 * F2FS_IPU_NOCACHE - disable IPU bio cache.
693 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
695 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
697 #define DEF_MIN_IPU_UTIL 70
698 #define DEF_MIN_FSYNC_BLOCKS 8
699 #define DEF_MIN_HOT_BLOCKS 16
701 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
711 F2FS_IPU_HONOR_OPU_WRITE,
714 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
717 struct curseg_info *curseg = CURSEG_I(sbi, type);
718 return curseg->segno;
721 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
724 struct curseg_info *curseg = CURSEG_I(sbi, type);
725 return curseg->alloc_type;
728 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
730 struct curseg_info *curseg = CURSEG_I(sbi, type);
731 return curseg->next_blkoff;
734 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
736 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
739 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
741 struct f2fs_sb_info *sbi = fio->sbi;
743 if (__is_valid_data_blkaddr(fio->old_blkaddr))
744 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
745 META_GENERIC : DATA_GENERIC);
746 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
747 META_GENERIC : DATA_GENERIC_ENHANCE);
751 * Summary block is always treated as an invalid block
753 static inline int check_block_count(struct f2fs_sb_info *sbi,
754 int segno, struct f2fs_sit_entry *raw_sit)
756 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
757 int valid_blocks = 0;
758 int cur_pos = 0, next_pos;
759 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
761 /* check bitmap with valid block count */
764 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
767 valid_blocks += next_pos - cur_pos;
769 next_pos = find_next_bit_le(&raw_sit->valid_map,
773 is_valid = !is_valid;
774 } while (cur_pos < usable_blks_per_seg);
776 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
777 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
778 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
779 set_sbi_flag(sbi, SBI_NEED_FSCK);
780 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
781 return -EFSCORRUPTED;
784 if (usable_blks_per_seg < sbi->blocks_per_seg)
785 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
787 usable_blks_per_seg) != sbi->blocks_per_seg);
789 /* check segment usage, and check boundary of a given segment number */
790 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
791 || segno > TOTAL_SEGS(sbi) - 1)) {
792 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
793 GET_SIT_VBLOCKS(raw_sit), segno);
794 set_sbi_flag(sbi, SBI_NEED_FSCK);
795 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
796 return -EFSCORRUPTED;
801 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
804 struct sit_info *sit_i = SIT_I(sbi);
805 unsigned int offset = SIT_BLOCK_OFFSET(start);
806 block_t blk_addr = sit_i->sit_base_addr + offset;
808 check_seg_range(sbi, start);
810 #ifdef CONFIG_F2FS_CHECK_FS
811 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
812 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
816 /* calculate sit block address */
817 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
818 blk_addr += sit_i->sit_blocks;
823 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
826 struct sit_info *sit_i = SIT_I(sbi);
827 block_addr -= sit_i->sit_base_addr;
828 if (block_addr < sit_i->sit_blocks)
829 block_addr += sit_i->sit_blocks;
831 block_addr -= sit_i->sit_blocks;
833 return block_addr + sit_i->sit_base_addr;
836 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
838 unsigned int block_off = SIT_BLOCK_OFFSET(start);
840 f2fs_change_bit(block_off, sit_i->sit_bitmap);
841 #ifdef CONFIG_F2FS_CHECK_FS
842 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
846 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
849 struct sit_info *sit_i = SIT_I(sbi);
850 time64_t diff, now = ktime_get_boottime_seconds();
852 if (now >= sit_i->mounted_time)
853 return sit_i->elapsed_time + now - sit_i->mounted_time;
855 /* system time is set to the past */
857 diff = sit_i->mounted_time - now;
858 if (sit_i->elapsed_time >= diff)
859 return sit_i->elapsed_time - diff;
862 return sit_i->elapsed_time;
865 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
866 unsigned int ofs_in_node, unsigned char version)
868 sum->nid = cpu_to_le32(nid);
869 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
870 sum->version = version;
873 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
875 return __start_cp_addr(sbi) +
876 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
879 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
881 return __start_cp_addr(sbi) +
882 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
886 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
888 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
894 * It is very important to gather dirty pages and write at once, so that we can
895 * submit a big bio without interfering other data writes.
896 * By default, 512 pages for directory data,
897 * 512 pages (2MB) * 8 for nodes, and
898 * 256 pages * 8 for meta are set.
900 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
902 if (sbi->sb->s_bdi->wb.dirty_exceeded)
906 return sbi->blocks_per_seg;
907 else if (type == NODE)
908 return 8 * sbi->blocks_per_seg;
909 else if (type == META)
910 return 8 * BIO_MAX_VECS;
916 * When writing pages, it'd better align nr_to_write for segment size.
918 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
919 struct writeback_control *wbc)
921 long nr_to_write, desired;
923 if (wbc->sync_mode != WB_SYNC_NONE)
926 nr_to_write = wbc->nr_to_write;
927 desired = BIO_MAX_VECS;
931 wbc->nr_to_write = desired;
932 return desired - nr_to_write;
935 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
937 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
944 mutex_lock(&dcc->cmd_lock);
945 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
946 if (i + 1 < dcc->discard_granularity)
948 if (!list_empty(&dcc->pend_list[i])) {
953 mutex_unlock(&dcc->cmd_lock);
954 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
957 dcc->discard_wake = 1;
958 wake_up_interruptible_all(&dcc->discard_wait_queue);