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 */
20 /* L: Logical segment # in volume, R: Relative segment # in main area */
21 #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
22 #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
24 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
25 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
27 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
28 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
29 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
31 #define IS_CURSEG(sbi, seg) \
32 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
33 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
34 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
35 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
36 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
37 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
39 #define IS_CURSEC(sbi, secno) \
40 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
41 (sbi)->segs_per_sec) || \
42 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
43 (sbi)->segs_per_sec) || \
44 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
45 (sbi)->segs_per_sec) || \
46 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
47 (sbi)->segs_per_sec) || \
48 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
49 (sbi)->segs_per_sec) || \
50 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
51 (sbi)->segs_per_sec)) \
53 #define MAIN_BLKADDR(sbi) \
54 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
55 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
56 #define SEG0_BLKADDR(sbi) \
57 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
58 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
60 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
61 #define MAIN_SECS(sbi) ((sbi)->total_sections)
63 #define TOTAL_SEGS(sbi) \
64 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
65 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
66 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
68 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
69 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
70 (sbi)->log_blocks_per_seg))
72 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
73 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
75 #define NEXT_FREE_BLKADDR(sbi, curseg) \
76 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
78 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
79 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
80 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
81 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
82 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
84 #define GET_SEGNO(sbi, blk_addr) \
85 ((!__is_valid_data_blkaddr(blk_addr)) ? \
86 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
87 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
88 #define BLKS_PER_SEC(sbi) \
89 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
90 #define GET_SEC_FROM_SEG(sbi, segno) \
91 ((segno) / (sbi)->segs_per_sec)
92 #define GET_SEG_FROM_SEC(sbi, secno) \
93 ((secno) * (sbi)->segs_per_sec)
94 #define GET_ZONE_FROM_SEC(sbi, secno) \
95 ((secno) / (sbi)->secs_per_zone)
96 #define GET_ZONE_FROM_SEG(sbi, segno) \
97 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
99 #define GET_SUM_BLOCK(sbi, segno) \
100 ((sbi)->sm_info->ssa_blkaddr + (segno))
102 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
103 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
105 #define SIT_ENTRY_OFFSET(sit_i, segno) \
106 ((segno) % (sit_i)->sents_per_block)
107 #define SIT_BLOCK_OFFSET(segno) \
108 ((segno) / SIT_ENTRY_PER_BLOCK)
109 #define START_SEGNO(segno) \
110 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
111 #define SIT_BLK_CNT(sbi) \
112 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
113 #define f2fs_bitmap_size(nr) \
114 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
116 #define SECTOR_FROM_BLOCK(blk_addr) \
117 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
118 #define SECTOR_TO_BLOCK(sectors) \
119 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
122 * indicate a block allocation direction: RIGHT and LEFT.
123 * RIGHT means allocating new sections towards the end of volume.
124 * LEFT means the opposite direction.
132 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
133 * LFS writes data sequentially with cleaning operations.
134 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
142 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
143 * GC_CB is based on cost-benefit algorithm.
144 * GC_GREEDY is based on greedy algorithm.
155 * BG_GC means the background cleaning job.
156 * FG_GC means the on-demand cleaning job.
157 * FORCE_FG_GC means on-demand cleaning job in background.
165 /* for a function parameter to select a victim segment */
166 struct victim_sel_policy {
167 int alloc_mode; /* LFS or SSR */
168 int gc_mode; /* GC_CB or GC_GREEDY */
169 unsigned long *dirty_segmap; /* dirty segment bitmap */
170 unsigned int max_search; /* maximum # of segments to search */
171 unsigned int offset; /* last scanned bitmap offset */
172 unsigned int ofs_unit; /* bitmap search unit */
173 unsigned int min_cost; /* minimum cost */
174 unsigned int min_segno; /* segment # having min. cost */
178 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
179 unsigned int valid_blocks:10; /* # of valid blocks */
180 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
181 unsigned int padding:6; /* padding */
182 unsigned char *cur_valid_map; /* validity bitmap of blocks */
183 #ifdef CONFIG_F2FS_CHECK_FS
184 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
187 * # of valid blocks and the validity bitmap stored in the the last
188 * checkpoint pack. This information is used by the SSR mode.
190 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
191 unsigned char *discard_map;
192 unsigned long long mtime; /* modification time of the segment */
196 unsigned int valid_blocks; /* # of valid blocks in a section */
199 struct segment_allocation {
200 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
204 * this value is set in page as a private data which indicate that
205 * the page is atomically written, and it is in inmem_pages list.
207 #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
208 #define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
210 #define IS_ATOMIC_WRITTEN_PAGE(page) \
211 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
212 #define IS_DUMMY_WRITTEN_PAGE(page) \
213 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
215 #define MAX_SKIP_GC_COUNT 16
218 struct list_head list;
220 block_t old_addr; /* for revoking when fail to commit */
224 const struct segment_allocation *s_ops;
226 block_t sit_base_addr; /* start block address of SIT area */
227 block_t sit_blocks; /* # of blocks used by SIT area */
228 block_t written_valid_blocks; /* # of valid blocks in main area */
229 char *bitmap; /* all bitmaps pointer */
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 && __is_large_section(sbi))
338 return get_sec_entry(sbi, segno)->valid_blocks;
340 return get_seg_entry(sbi, segno)->valid_blocks;
343 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
346 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
349 static inline void seg_info_from_raw_sit(struct seg_entry *se,
350 struct f2fs_sit_entry *rs)
352 se->valid_blocks = GET_SIT_VBLOCKS(rs);
353 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
354 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
355 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
356 #ifdef CONFIG_F2FS_CHECK_FS
357 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
359 se->type = GET_SIT_TYPE(rs);
360 se->mtime = le64_to_cpu(rs->mtime);
363 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
364 struct f2fs_sit_entry *rs)
366 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
368 rs->vblocks = cpu_to_le16(raw_vblocks);
369 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
370 rs->mtime = cpu_to_le64(se->mtime);
373 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
374 struct page *page, unsigned int start)
376 struct f2fs_sit_block *raw_sit;
377 struct seg_entry *se;
378 struct f2fs_sit_entry *rs;
379 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
380 (unsigned long)MAIN_SEGS(sbi));
383 raw_sit = (struct f2fs_sit_block *)page_address(page);
384 memset(raw_sit, 0, PAGE_SIZE);
385 for (i = 0; i < end - start; i++) {
386 rs = &raw_sit->entries[i];
387 se = get_seg_entry(sbi, start + i);
388 __seg_info_to_raw_sit(se, rs);
392 static inline void seg_info_to_raw_sit(struct seg_entry *se,
393 struct f2fs_sit_entry *rs)
395 __seg_info_to_raw_sit(se, rs);
397 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
398 se->ckpt_valid_blocks = se->valid_blocks;
401 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
402 unsigned int max, unsigned int segno)
405 spin_lock(&free_i->segmap_lock);
406 ret = find_next_bit(free_i->free_segmap, max, segno);
407 spin_unlock(&free_i->segmap_lock);
411 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
413 struct free_segmap_info *free_i = FREE_I(sbi);
414 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
415 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
418 spin_lock(&free_i->segmap_lock);
419 clear_bit(segno, free_i->free_segmap);
420 free_i->free_segments++;
422 next = find_next_bit(free_i->free_segmap,
423 start_segno + sbi->segs_per_sec, start_segno);
424 if (next >= start_segno + sbi->segs_per_sec) {
425 clear_bit(secno, free_i->free_secmap);
426 free_i->free_sections++;
428 spin_unlock(&free_i->segmap_lock);
431 static inline void __set_inuse(struct f2fs_sb_info *sbi,
434 struct free_segmap_info *free_i = FREE_I(sbi);
435 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
437 set_bit(segno, free_i->free_segmap);
438 free_i->free_segments--;
439 if (!test_and_set_bit(secno, free_i->free_secmap))
440 free_i->free_sections--;
443 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
446 struct free_segmap_info *free_i = FREE_I(sbi);
447 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
448 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
451 spin_lock(&free_i->segmap_lock);
452 if (test_and_clear_bit(segno, free_i->free_segmap)) {
453 free_i->free_segments++;
455 if (IS_CURSEC(sbi, secno))
457 next = find_next_bit(free_i->free_segmap,
458 start_segno + sbi->segs_per_sec, start_segno);
459 if (next >= start_segno + sbi->segs_per_sec) {
460 if (test_and_clear_bit(secno, free_i->free_secmap))
461 free_i->free_sections++;
465 spin_unlock(&free_i->segmap_lock);
468 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
471 struct free_segmap_info *free_i = FREE_I(sbi);
472 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
474 spin_lock(&free_i->segmap_lock);
475 if (!test_and_set_bit(segno, free_i->free_segmap)) {
476 free_i->free_segments--;
477 if (!test_and_set_bit(secno, free_i->free_secmap))
478 free_i->free_sections--;
480 spin_unlock(&free_i->segmap_lock);
483 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
486 struct sit_info *sit_i = SIT_I(sbi);
488 #ifdef CONFIG_F2FS_CHECK_FS
489 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
493 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
496 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
498 return SIT_I(sbi)->written_valid_blocks;
501 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
503 return FREE_I(sbi)->free_segments;
506 static inline int reserved_segments(struct f2fs_sb_info *sbi)
508 return SM_I(sbi)->reserved_segments;
511 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
513 return FREE_I(sbi)->free_sections;
516 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
518 return DIRTY_I(sbi)->nr_dirty[PRE];
521 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
523 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
524 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
525 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
526 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
527 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
528 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
531 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
533 return SM_I(sbi)->ovp_segments;
536 static inline int reserved_sections(struct f2fs_sb_info *sbi)
538 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
541 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
543 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
544 get_pages(sbi, F2FS_DIRTY_DENTS);
545 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
546 unsigned int segno, left_blocks;
549 /* check current node segment */
550 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
551 segno = CURSEG_I(sbi, i)->segno;
552 left_blocks = sbi->blocks_per_seg -
553 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
555 if (node_blocks > left_blocks)
559 /* check current data segment */
560 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
561 left_blocks = sbi->blocks_per_seg -
562 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
563 if (dent_blocks > left_blocks)
568 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
569 int freed, int needed)
571 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
572 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
573 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
575 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
578 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
579 has_curseg_enough_space(sbi))
581 return (free_sections(sbi) + freed) <=
582 (node_secs + 2 * dent_secs + imeta_secs +
583 reserved_sections(sbi) + needed);
586 static inline int f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
588 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
590 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
595 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
597 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
600 static inline int utilization(struct f2fs_sb_info *sbi)
602 return div_u64((u64)valid_user_blocks(sbi) * 100,
603 sbi->user_block_count);
607 * Sometimes f2fs may be better to drop out-of-place update policy.
608 * And, users can control the policy through sysfs entries.
609 * There are five policies with triggering conditions as follows.
610 * F2FS_IPU_FORCE - all the time,
611 * F2FS_IPU_SSR - if SSR mode is activated,
612 * F2FS_IPU_UTIL - if FS utilization is over threashold,
613 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
615 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
616 * storages. IPU will be triggered only if the # of dirty
617 * pages over min_fsync_blocks.
618 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
620 #define DEF_MIN_IPU_UTIL 70
621 #define DEF_MIN_FSYNC_BLOCKS 8
622 #define DEF_MIN_HOT_BLOCKS 16
624 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
635 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
638 struct curseg_info *curseg = CURSEG_I(sbi, type);
639 return curseg->segno;
642 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
645 struct curseg_info *curseg = CURSEG_I(sbi, type);
646 return curseg->alloc_type;
649 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
651 struct curseg_info *curseg = CURSEG_I(sbi, type);
652 return curseg->next_blkoff;
655 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
657 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
660 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
662 struct f2fs_sb_info *sbi = fio->sbi;
664 if (__is_valid_data_blkaddr(fio->old_blkaddr))
665 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
666 META_GENERIC : DATA_GENERIC);
667 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
668 META_GENERIC : DATA_GENERIC_ENHANCE);
672 * Summary block is always treated as an invalid block
674 static inline int check_block_count(struct f2fs_sb_info *sbi,
675 int segno, struct f2fs_sit_entry *raw_sit)
677 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
678 int valid_blocks = 0;
679 int cur_pos = 0, next_pos;
681 /* check bitmap with valid block count */
684 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
687 valid_blocks += next_pos - cur_pos;
689 next_pos = find_next_bit_le(&raw_sit->valid_map,
693 is_valid = !is_valid;
694 } while (cur_pos < sbi->blocks_per_seg);
696 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
697 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
698 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
699 set_sbi_flag(sbi, SBI_NEED_FSCK);
700 return -EFSCORRUPTED;
703 /* check segment usage, and check boundary of a given segment number */
704 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
705 || segno > TOTAL_SEGS(sbi) - 1)) {
706 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
707 GET_SIT_VBLOCKS(raw_sit), segno);
708 set_sbi_flag(sbi, SBI_NEED_FSCK);
709 return -EFSCORRUPTED;
714 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
717 struct sit_info *sit_i = SIT_I(sbi);
718 unsigned int offset = SIT_BLOCK_OFFSET(start);
719 block_t blk_addr = sit_i->sit_base_addr + offset;
721 check_seg_range(sbi, start);
723 #ifdef CONFIG_F2FS_CHECK_FS
724 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
725 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
729 /* calculate sit block address */
730 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
731 blk_addr += sit_i->sit_blocks;
736 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
739 struct sit_info *sit_i = SIT_I(sbi);
740 block_addr -= sit_i->sit_base_addr;
741 if (block_addr < sit_i->sit_blocks)
742 block_addr += sit_i->sit_blocks;
744 block_addr -= sit_i->sit_blocks;
746 return block_addr + sit_i->sit_base_addr;
749 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
751 unsigned int block_off = SIT_BLOCK_OFFSET(start);
753 f2fs_change_bit(block_off, sit_i->sit_bitmap);
754 #ifdef CONFIG_F2FS_CHECK_FS
755 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
759 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
762 struct sit_info *sit_i = SIT_I(sbi);
763 time64_t diff, now = ktime_get_real_seconds();
765 if (now >= sit_i->mounted_time)
766 return sit_i->elapsed_time + now - sit_i->mounted_time;
768 /* system time is set to the past */
770 diff = sit_i->mounted_time - now;
771 if (sit_i->elapsed_time >= diff)
772 return sit_i->elapsed_time - diff;
775 return sit_i->elapsed_time;
778 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
779 unsigned int ofs_in_node, unsigned char version)
781 sum->nid = cpu_to_le32(nid);
782 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
783 sum->version = version;
786 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
788 return __start_cp_addr(sbi) +
789 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
792 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
794 return __start_cp_addr(sbi) +
795 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
799 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
801 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
807 * It is very important to gather dirty pages and write at once, so that we can
808 * submit a big bio without interfering other data writes.
809 * By default, 512 pages for directory data,
810 * 512 pages (2MB) * 8 for nodes, and
811 * 256 pages * 8 for meta are set.
813 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
815 if (sbi->sb->s_bdi->wb.dirty_exceeded)
819 return sbi->blocks_per_seg;
820 else if (type == NODE)
821 return 8 * sbi->blocks_per_seg;
822 else if (type == META)
823 return 8 * BIO_MAX_PAGES;
829 * When writing pages, it'd better align nr_to_write for segment size.
831 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
832 struct writeback_control *wbc)
834 long nr_to_write, desired;
836 if (wbc->sync_mode != WB_SYNC_NONE)
839 nr_to_write = wbc->nr_to_write;
840 desired = BIO_MAX_PAGES;
844 wbc->nr_to_write = desired;
845 return desired - nr_to_write;
848 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
850 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
857 mutex_lock(&dcc->cmd_lock);
858 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
859 if (i + 1 < dcc->discard_granularity)
861 if (!list_empty(&dcc->pend_list[i])) {
866 mutex_unlock(&dcc->cmd_lock);
867 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
870 dcc->discard_wake = 1;
871 wake_up_interruptible_all(&dcc->discard_wait_queue);