4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
11 #include <linux/blkdev.h>
12 #include <linux/backing-dev.h>
15 #define NULL_SEGNO ((unsigned int)(~0))
16 #define NULL_SECNO ((unsigned int)(~0))
18 #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
19 #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
21 #define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
23 /* L: Logical segment # in volume, R: Relative segment # in main area */
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)
27 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
28 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
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)
34 #define IS_CURSEG(sbi, seg) \
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))
42 #define IS_CURSEC(sbi, secno) \
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)) \
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))
63 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
64 #define MAIN_SECS(sbi) ((sbi)->total_sections)
66 #define TOTAL_SEGS(sbi) \
67 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
68 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
69 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
71 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
72 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
73 (sbi)->log_blocks_per_seg))
75 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
76 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
78 #define NEXT_FREE_BLKADDR(sbi, curseg) \
79 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
81 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
82 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
83 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
84 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
85 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
87 #define GET_SEGNO(sbi, blk_addr) \
88 ((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ? \
89 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
90 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
91 #define BLKS_PER_SEC(sbi) \
92 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
93 #define GET_SEC_FROM_SEG(sbi, segno) \
94 ((segno) / (sbi)->segs_per_sec)
95 #define GET_SEG_FROM_SEC(sbi, secno) \
96 ((secno) * (sbi)->segs_per_sec)
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))
102 #define GET_SUM_BLOCK(sbi, segno) \
103 ((sbi)->sm_info->ssa_blkaddr + (segno))
105 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
106 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
108 #define SIT_ENTRY_OFFSET(sit_i, segno) \
109 ((segno) % (sit_i)->sents_per_block)
110 #define SIT_BLOCK_OFFSET(segno) \
111 ((segno) / SIT_ENTRY_PER_BLOCK)
112 #define START_SEGNO(segno) \
113 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
114 #define SIT_BLK_CNT(sbi) \
115 ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
116 #define f2fs_bitmap_size(nr) \
117 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
119 #define SECTOR_FROM_BLOCK(blk_addr) \
120 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
121 #define SECTOR_TO_BLOCK(sectors) \
122 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
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.
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.
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.
158 * BG_GC means the background cleaning job.
159 * FG_GC means the on-demand cleaning job.
160 * FORCE_FG_GC means on-demand cleaning job in background.
168 /* for a function parameter to select a victim segment */
169 struct 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 */
173 unsigned int max_search; /* maximum # of segments to search */
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 */
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 */
185 unsigned char *cur_valid_map; /* validity bitmap of blocks */
186 #ifdef CONFIG_F2FS_CHECK_FS
187 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
190 * # of valid blocks and the validity bitmap stored in the the last
191 * checkpoint pack. This information is used by the SSR mode.
193 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
194 unsigned char *discard_map;
195 unsigned long long mtime; /* modification time of the segment */
199 unsigned int valid_blocks; /* # of valid blocks in a section */
202 struct segment_allocation {
203 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
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.
210 #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
211 #define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
213 #define IS_ATOMIC_WRITTEN_PAGE(page) \
214 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
215 #define IS_DUMMY_WRITTEN_PAGE(page) \
216 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
219 struct list_head list;
221 block_t old_addr; /* for revoking when fail to commit */
225 const struct segment_allocation *s_ops;
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 */
231 #ifdef CONFIG_F2FS_CHECK_FS
232 char *sit_bitmap_mir; /* SIT bitmap mirror */
234 unsigned int bitmap_size; /* SIT bitmap size */
236 unsigned long *tmp_map; /* bitmap for temporal use */
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 */
240 struct rw_semaphore sentry_lock; /* to protect SIT cache */
241 struct seg_entry *sentries; /* SIT segment-level cache */
242 struct sec_entry *sec_entries; /* SIT section-level cache */
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 */
250 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
253 struct 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 */
257 spinlock_t segmap_lock; /* free segmap lock */
258 unsigned long *free_segmap; /* free segment bitmap */
259 unsigned long *free_secmap; /* free section bitmap */
262 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
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 */
275 struct 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 */
280 unsigned long *victim_secmap; /* background GC victims */
283 /* victim selection function for cleaning and SSR */
284 struct victim_selection {
285 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
289 /* for active log information */
291 struct mutex curseg_mutex; /* lock for consistency */
292 struct f2fs_summary_block *sum_blk; /* cached summary block */
293 struct rw_semaphore journal_rwsem; /* protect journal area */
294 struct f2fs_journal *journal; /* cached journal info */
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 */
302 struct 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 */
311 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
313 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
316 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
319 struct sit_info *sit_i = SIT_I(sbi);
320 return &sit_i->sentries[segno];
323 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
326 struct sit_info *sit_i = SIT_I(sbi);
327 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
330 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
331 unsigned int segno, bool use_section)
334 * In order to get # of valid blocks in a section instantly from many
335 * segments, f2fs manages two counting structures separately.
337 if (use_section && sbi->segs_per_sec > 1)
338 return get_sec_entry(sbi, segno)->valid_blocks;
340 return get_seg_entry(sbi, segno)->valid_blocks;
343 static inline void seg_info_from_raw_sit(struct seg_entry *se,
344 struct f2fs_sit_entry *rs)
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);
350 #ifdef CONFIG_F2FS_CHECK_FS
351 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
353 se->type = GET_SIT_TYPE(rs);
354 se->mtime = le64_to_cpu(rs->mtime);
357 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
358 struct f2fs_sit_entry *rs)
360 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
362 rs->vblocks = cpu_to_le16(raw_vblocks);
363 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
364 rs->mtime = cpu_to_le64(se->mtime);
367 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
368 struct page *page, unsigned int start)
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));
377 raw_sit = (struct f2fs_sit_block *)page_address(page);
378 for (i = 0; i < end - start; i++) {
379 rs = &raw_sit->entries[i];
380 se = get_seg_entry(sbi, start + i);
381 __seg_info_to_raw_sit(se, rs);
385 static inline void seg_info_to_raw_sit(struct seg_entry *se,
386 struct f2fs_sit_entry *rs)
388 __seg_info_to_raw_sit(se, rs);
390 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
391 se->ckpt_valid_blocks = se->valid_blocks;
394 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
395 unsigned int max, unsigned int segno)
398 spin_lock(&free_i->segmap_lock);
399 ret = find_next_bit(free_i->free_segmap, max, segno);
400 spin_unlock(&free_i->segmap_lock);
404 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
406 struct free_segmap_info *free_i = FREE_I(sbi);
407 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
408 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
411 spin_lock(&free_i->segmap_lock);
412 clear_bit(segno, free_i->free_segmap);
413 free_i->free_segments++;
415 next = find_next_bit(free_i->free_segmap,
416 start_segno + sbi->segs_per_sec, start_segno);
417 if (next >= start_segno + sbi->segs_per_sec) {
418 clear_bit(secno, free_i->free_secmap);
419 free_i->free_sections++;
421 spin_unlock(&free_i->segmap_lock);
424 static inline void __set_inuse(struct f2fs_sb_info *sbi,
427 struct free_segmap_info *free_i = FREE_I(sbi);
428 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
430 set_bit(segno, free_i->free_segmap);
431 free_i->free_segments--;
432 if (!test_and_set_bit(secno, free_i->free_secmap))
433 free_i->free_sections--;
436 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
439 struct free_segmap_info *free_i = FREE_I(sbi);
440 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
441 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
444 spin_lock(&free_i->segmap_lock);
445 if (test_and_clear_bit(segno, free_i->free_segmap)) {
446 free_i->free_segments++;
448 next = find_next_bit(free_i->free_segmap,
449 start_segno + sbi->segs_per_sec, start_segno);
450 if (next >= start_segno + sbi->segs_per_sec) {
451 if (test_and_clear_bit(secno, free_i->free_secmap))
452 free_i->free_sections++;
455 spin_unlock(&free_i->segmap_lock);
458 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
461 struct free_segmap_info *free_i = FREE_I(sbi);
462 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
464 spin_lock(&free_i->segmap_lock);
465 if (!test_and_set_bit(segno, free_i->free_segmap)) {
466 free_i->free_segments--;
467 if (!test_and_set_bit(secno, free_i->free_secmap))
468 free_i->free_sections--;
470 spin_unlock(&free_i->segmap_lock);
473 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
476 struct sit_info *sit_i = SIT_I(sbi);
478 #ifdef CONFIG_F2FS_CHECK_FS
479 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
483 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
486 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
488 return SIT_I(sbi)->written_valid_blocks;
491 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
493 return FREE_I(sbi)->free_segments;
496 static inline int reserved_segments(struct f2fs_sb_info *sbi)
498 return SM_I(sbi)->reserved_segments;
501 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
503 return FREE_I(sbi)->free_sections;
506 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
508 return DIRTY_I(sbi)->nr_dirty[PRE];
511 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
513 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
514 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
515 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
516 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
517 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
518 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
521 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
523 return SM_I(sbi)->ovp_segments;
526 static inline int reserved_sections(struct f2fs_sb_info *sbi)
528 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
531 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
533 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
534 get_pages(sbi, F2FS_DIRTY_DENTS);
535 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
536 unsigned int segno, left_blocks;
539 /* check current node segment */
540 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
541 segno = CURSEG_I(sbi, i)->segno;
542 left_blocks = sbi->blocks_per_seg -
543 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
545 if (node_blocks > left_blocks)
549 /* check current data segment */
550 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
551 left_blocks = sbi->blocks_per_seg -
552 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
553 if (dent_blocks > left_blocks)
558 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
559 int freed, int needed)
561 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
562 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
563 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
565 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
568 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
569 has_curseg_enough_space(sbi))
571 return (free_sections(sbi) + freed) <=
572 (node_secs + 2 * dent_secs + imeta_secs +
573 reserved_sections(sbi) + needed);
576 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
578 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
581 static inline int utilization(struct f2fs_sb_info *sbi)
583 return div_u64((u64)valid_user_blocks(sbi) * 100,
584 sbi->user_block_count);
588 * Sometimes f2fs may be better to drop out-of-place update policy.
589 * And, users can control the policy through sysfs entries.
590 * There are five policies with triggering conditions as follows.
591 * F2FS_IPU_FORCE - all the time,
592 * F2FS_IPU_SSR - if SSR mode is activated,
593 * F2FS_IPU_UTIL - if FS utilization is over threashold,
594 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
596 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
597 * storages. IPU will be triggered only if the # of dirty
598 * pages over min_fsync_blocks.
599 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
601 #define DEF_MIN_IPU_UTIL 70
602 #define DEF_MIN_FSYNC_BLOCKS 8
603 #define DEF_MIN_HOT_BLOCKS 16
605 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
616 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
619 struct curseg_info *curseg = CURSEG_I(sbi, type);
620 return curseg->segno;
623 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
626 struct curseg_info *curseg = CURSEG_I(sbi, type);
627 return curseg->alloc_type;
630 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
632 struct curseg_info *curseg = CURSEG_I(sbi, type);
633 return curseg->next_blkoff;
636 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
638 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
641 static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
643 struct f2fs_sb_info *sbi = fio->sbi;
645 if (PAGE_TYPE_OF_BIO(fio->type) == META &&
646 (!is_read_io(fio->op) || fio->is_meta))
647 BUG_ON(blk_addr < SEG0_BLKADDR(sbi) ||
648 blk_addr >= MAIN_BLKADDR(sbi));
650 BUG_ON(blk_addr < MAIN_BLKADDR(sbi) ||
651 blk_addr >= MAX_BLKADDR(sbi));
655 * Summary block is always treated as an invalid block
657 static inline int check_block_count(struct f2fs_sb_info *sbi,
658 int segno, struct f2fs_sit_entry *raw_sit)
660 #ifdef CONFIG_F2FS_CHECK_FS
661 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
662 int valid_blocks = 0;
663 int cur_pos = 0, next_pos;
665 /* check bitmap with valid block count */
668 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
671 valid_blocks += next_pos - cur_pos;
673 next_pos = find_next_bit_le(&raw_sit->valid_map,
677 is_valid = !is_valid;
678 } while (cur_pos < sbi->blocks_per_seg);
680 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
681 f2fs_msg(sbi->sb, KERN_ERR,
682 "Mismatch valid blocks %d vs. %d",
683 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
684 set_sbi_flag(sbi, SBI_NEED_FSCK);
688 /* check segment usage, and check boundary of a given segment number */
689 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
690 || segno > TOTAL_SEGS(sbi) - 1)) {
691 f2fs_msg(sbi->sb, KERN_ERR,
692 "Wrong valid blocks %d or segno %u",
693 GET_SIT_VBLOCKS(raw_sit), segno);
694 set_sbi_flag(sbi, SBI_NEED_FSCK);
700 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
703 struct sit_info *sit_i = SIT_I(sbi);
704 unsigned int offset = SIT_BLOCK_OFFSET(start);
705 block_t blk_addr = sit_i->sit_base_addr + offset;
707 check_seg_range(sbi, start);
709 #ifdef CONFIG_F2FS_CHECK_FS
710 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
711 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
715 /* calculate sit block address */
716 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
717 blk_addr += sit_i->sit_blocks;
722 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
725 struct sit_info *sit_i = SIT_I(sbi);
726 block_addr -= sit_i->sit_base_addr;
727 if (block_addr < sit_i->sit_blocks)
728 block_addr += sit_i->sit_blocks;
730 block_addr -= sit_i->sit_blocks;
732 return block_addr + sit_i->sit_base_addr;
735 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
737 unsigned int block_off = SIT_BLOCK_OFFSET(start);
739 f2fs_change_bit(block_off, sit_i->sit_bitmap);
740 #ifdef CONFIG_F2FS_CHECK_FS
741 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
745 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
747 struct sit_info *sit_i = SIT_I(sbi);
748 time64_t now = ktime_get_real_seconds();
750 return sit_i->elapsed_time + now - sit_i->mounted_time;
753 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
754 unsigned int ofs_in_node, unsigned char version)
756 sum->nid = cpu_to_le32(nid);
757 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
758 sum->version = version;
761 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
763 return __start_cp_addr(sbi) +
764 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
767 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
769 return __start_cp_addr(sbi) +
770 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
774 static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
777 if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >
783 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
785 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
791 * It is very important to gather dirty pages and write at once, so that we can
792 * submit a big bio without interfering other data writes.
793 * By default, 512 pages for directory data,
794 * 512 pages (2MB) * 8 for nodes, and
795 * 256 pages * 8 for meta are set.
797 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
799 if (sbi->sb->s_bdi->wb.dirty_exceeded)
803 return sbi->blocks_per_seg;
804 else if (type == NODE)
805 return 8 * sbi->blocks_per_seg;
806 else if (type == META)
807 return 8 * BIO_MAX_PAGES;
813 * When writing pages, it'd better align nr_to_write for segment size.
815 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
816 struct writeback_control *wbc)
818 long nr_to_write, desired;
820 if (wbc->sync_mode != WB_SYNC_NONE)
823 nr_to_write = wbc->nr_to_write;
824 desired = BIO_MAX_PAGES;
828 wbc->nr_to_write = desired;
829 return desired - nr_to_write;
832 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
834 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
841 mutex_lock(&dcc->cmd_lock);
842 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
843 if (i + 1 < dcc->discard_granularity)
845 if (!list_empty(&dcc->pend_list[i])) {
850 mutex_unlock(&dcc->cmd_lock);
854 dcc->discard_wake = 1;
855 wake_up_interruptible_all(&dcc->discard_wait_queue);