Loading...
1/*
2 * fs/f2fs/segment.h
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
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.
10 */
11#include <linux/blkdev.h>
12#include <linux/backing-dev.h>
13
14/* constant macro */
15#define NULL_SEGNO ((unsigned int)(~0))
16#define NULL_SECNO ((unsigned int)(~0))
17
18#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
19#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
20
21#define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
22
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)
26
27#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
28#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
29
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)
33
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))
41
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)) \
55
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))
62
63#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
64#define MAIN_SECS(sbi) ((sbi)->total_sections)
65
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)
70
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))
74
75#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
76 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
77
78#define NEXT_FREE_BLKADDR(sbi, curseg) \
79 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
80
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))
86
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))
101
102#define GET_SUM_BLOCK(sbi, segno) \
103 ((sbi)->sm_info->ssa_blkaddr + (segno))
104
105#define GET_SUM_TYPE(footer) ((footer)->entry_type)
106#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
107
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))
118
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)
123
124/*
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.
128 */
129enum {
130 ALLOC_RIGHT = 0,
131 ALLOC_LEFT
132};
133
134/*
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.
138 */
139enum {
140 LFS = 0,
141 SSR
142};
143
144/*
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.
148 */
149enum {
150 GC_CB = 0,
151 GC_GREEDY,
152 ALLOC_NEXT,
153 FLUSH_DEVICE,
154 MAX_GC_POLICY,
155};
156
157/*
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.
161 */
162enum {
163 BG_GC = 0,
164 FG_GC,
165 FORCE_FG_GC,
166};
167
168/* for a function parameter to select a victim segment */
169struct 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 */
178};
179
180struct seg_entry {
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 */
188#endif
189 /*
190 * # of valid blocks and the validity bitmap stored in the the last
191 * checkpoint pack. This information is used by the SSR mode.
192 */
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 */
196};
197
198struct sec_entry {
199 unsigned int valid_blocks; /* # of valid blocks in a section */
200};
201
202struct segment_allocation {
203 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
204};
205
206/*
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.
209 */
210#define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
211#define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
212
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)
217
218struct inmem_pages {
219 struct list_head list;
220 struct page *page;
221 block_t old_addr; /* for revoking when fail to commit */
222};
223
224struct sit_info {
225 const struct segment_allocation *s_ops;
226
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 */
233#endif
234 unsigned int bitmap_size; /* SIT bitmap size */
235
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 */
243
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 */
249
250 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
251};
252
253struct 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 */
260};
261
262/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
263enum dirty_type {
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 */
272 NR_DIRTY_TYPE
273};
274
275struct 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 */
281};
282
283/* victim selection function for cleaning and SSR */
284struct victim_selection {
285 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
286 int, int, char);
287};
288
289/* for active log information */
290struct curseg_info {
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 */
300};
301
302struct 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 */
306};
307
308/*
309 * inline functions
310 */
311static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
312{
313 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
314}
315
316static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
317 unsigned int segno)
318{
319 struct sit_info *sit_i = SIT_I(sbi);
320 return &sit_i->sentries[segno];
321}
322
323static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
324 unsigned int segno)
325{
326 struct sit_info *sit_i = SIT_I(sbi);
327 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
328}
329
330static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
331 unsigned int segno, bool use_section)
332{
333 /*
334 * In order to get # of valid blocks in a section instantly from many
335 * segments, f2fs manages two counting structures separately.
336 */
337 if (use_section && sbi->segs_per_sec > 1)
338 return get_sec_entry(sbi, segno)->valid_blocks;
339 else
340 return get_seg_entry(sbi, segno)->valid_blocks;
341}
342
343static inline void seg_info_from_raw_sit(struct seg_entry *se,
344 struct f2fs_sit_entry *rs)
345{
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);
352#endif
353 se->type = GET_SIT_TYPE(rs);
354 se->mtime = le64_to_cpu(rs->mtime);
355}
356
357static inline void __seg_info_to_raw_sit(struct seg_entry *se,
358 struct f2fs_sit_entry *rs)
359{
360 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
361 se->valid_blocks;
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);
365}
366
367static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
368 struct page *page, unsigned int start)
369{
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));
375 int i;
376
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);
382 }
383}
384
385static inline void seg_info_to_raw_sit(struct seg_entry *se,
386 struct f2fs_sit_entry *rs)
387{
388 __seg_info_to_raw_sit(se, rs);
389
390 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
391 se->ckpt_valid_blocks = se->valid_blocks;
392}
393
394static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
395 unsigned int max, unsigned int segno)
396{
397 unsigned int ret;
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);
401 return ret;
402}
403
404static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
405{
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);
409 unsigned int next;
410
411 spin_lock(&free_i->segmap_lock);
412 clear_bit(segno, free_i->free_segmap);
413 free_i->free_segments++;
414
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++;
420 }
421 spin_unlock(&free_i->segmap_lock);
422}
423
424static inline void __set_inuse(struct f2fs_sb_info *sbi,
425 unsigned int segno)
426{
427 struct free_segmap_info *free_i = FREE_I(sbi);
428 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
429
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--;
434}
435
436static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
437 unsigned int segno)
438{
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);
442 unsigned int next;
443
444 spin_lock(&free_i->segmap_lock);
445 if (test_and_clear_bit(segno, free_i->free_segmap)) {
446 free_i->free_segments++;
447
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++;
453 }
454 }
455 spin_unlock(&free_i->segmap_lock);
456}
457
458static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
459 unsigned int segno)
460{
461 struct free_segmap_info *free_i = FREE_I(sbi);
462 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
463
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--;
469 }
470 spin_unlock(&free_i->segmap_lock);
471}
472
473static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
474 void *dst_addr)
475{
476 struct sit_info *sit_i = SIT_I(sbi);
477
478#ifdef CONFIG_F2FS_CHECK_FS
479 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
480 sit_i->bitmap_size))
481 f2fs_bug_on(sbi, 1);
482#endif
483 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
484}
485
486static inline block_t written_block_count(struct f2fs_sb_info *sbi)
487{
488 return SIT_I(sbi)->written_valid_blocks;
489}
490
491static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
492{
493 return FREE_I(sbi)->free_segments;
494}
495
496static inline int reserved_segments(struct f2fs_sb_info *sbi)
497{
498 return SM_I(sbi)->reserved_segments;
499}
500
501static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
502{
503 return FREE_I(sbi)->free_sections;
504}
505
506static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
507{
508 return DIRTY_I(sbi)->nr_dirty[PRE];
509}
510
511static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
512{
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];
519}
520
521static inline int overprovision_segments(struct f2fs_sb_info *sbi)
522{
523 return SM_I(sbi)->ovp_segments;
524}
525
526static inline int reserved_sections(struct f2fs_sb_info *sbi)
527{
528 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
529}
530
531static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
532{
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;
537 int i;
538
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;
544
545 if (node_blocks > left_blocks)
546 return false;
547 }
548
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)
554 return false;
555 return true;
556}
557
558static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
559 int freed, int needed)
560{
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);
564
565 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
566 return false;
567
568 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
569 has_curseg_enough_space(sbi))
570 return false;
571 return (free_sections(sbi) + freed) <=
572 (node_secs + 2 * dent_secs + imeta_secs +
573 reserved_sections(sbi) + needed);
574}
575
576static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
577{
578 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
579}
580
581static inline int utilization(struct f2fs_sb_info *sbi)
582{
583 return div_u64((u64)valid_user_blocks(sbi) * 100,
584 sbi->user_block_count);
585}
586
587/*
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
595 * threashold,
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)
600 */
601#define DEF_MIN_IPU_UTIL 70
602#define DEF_MIN_FSYNC_BLOCKS 8
603#define DEF_MIN_HOT_BLOCKS 16
604
605#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
606
607enum {
608 F2FS_IPU_FORCE,
609 F2FS_IPU_SSR,
610 F2FS_IPU_UTIL,
611 F2FS_IPU_SSR_UTIL,
612 F2FS_IPU_FSYNC,
613 F2FS_IPU_ASYNC,
614};
615
616static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
617 int type)
618{
619 struct curseg_info *curseg = CURSEG_I(sbi, type);
620 return curseg->segno;
621}
622
623static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
624 int type)
625{
626 struct curseg_info *curseg = CURSEG_I(sbi, type);
627 return curseg->alloc_type;
628}
629
630static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
631{
632 struct curseg_info *curseg = CURSEG_I(sbi, type);
633 return curseg->next_blkoff;
634}
635
636static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
637{
638 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
639}
640
641static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
642{
643 struct f2fs_sb_info *sbi = fio->sbi;
644
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));
649 else
650 BUG_ON(blk_addr < MAIN_BLKADDR(sbi) ||
651 blk_addr >= MAX_BLKADDR(sbi));
652}
653
654/*
655 * Summary block is always treated as an invalid block
656 */
657static inline int check_block_count(struct f2fs_sb_info *sbi,
658 int segno, struct f2fs_sit_entry *raw_sit)
659{
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;
664
665 /* check bitmap with valid block count */
666 do {
667 if (is_valid) {
668 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
669 sbi->blocks_per_seg,
670 cur_pos);
671 valid_blocks += next_pos - cur_pos;
672 } else
673 next_pos = find_next_bit_le(&raw_sit->valid_map,
674 sbi->blocks_per_seg,
675 cur_pos);
676 cur_pos = next_pos;
677 is_valid = !is_valid;
678 } while (cur_pos < sbi->blocks_per_seg);
679
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);
685 return -EINVAL;
686 }
687#endif
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);
695 return -EINVAL;
696 }
697 return 0;
698}
699
700static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
701 unsigned int start)
702{
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;
706
707 check_seg_range(sbi, start);
708
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))
712 f2fs_bug_on(sbi, 1);
713#endif
714
715 /* calculate sit block address */
716 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
717 blk_addr += sit_i->sit_blocks;
718
719 return blk_addr;
720}
721
722static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
723 pgoff_t block_addr)
724{
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;
729 else
730 block_addr -= sit_i->sit_blocks;
731
732 return block_addr + sit_i->sit_base_addr;
733}
734
735static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
736{
737 unsigned int block_off = SIT_BLOCK_OFFSET(start);
738
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);
742#endif
743}
744
745static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
746{
747 struct sit_info *sit_i = SIT_I(sbi);
748 time64_t now = ktime_get_real_seconds();
749
750 return sit_i->elapsed_time + now - sit_i->mounted_time;
751}
752
753static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
754 unsigned int ofs_in_node, unsigned char version)
755{
756 sum->nid = cpu_to_le32(nid);
757 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
758 sum->version = version;
759}
760
761static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
762{
763 return __start_cp_addr(sbi) +
764 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
765}
766
767static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
768{
769 return __start_cp_addr(sbi) +
770 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
771 - (base + 1) + type;
772}
773
774static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
775 unsigned int secno)
776{
777 if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >
778 sbi->fggc_threshold)
779 return true;
780 return false;
781}
782
783static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
784{
785 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
786 return true;
787 return false;
788}
789
790/*
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.
796 */
797static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
798{
799 if (sbi->sb->s_bdi->wb.dirty_exceeded)
800 return 0;
801
802 if (type == DATA)
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;
808 else
809 return 0;
810}
811
812/*
813 * When writing pages, it'd better align nr_to_write for segment size.
814 */
815static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
816 struct writeback_control *wbc)
817{
818 long nr_to_write, desired;
819
820 if (wbc->sync_mode != WB_SYNC_NONE)
821 return 0;
822
823 nr_to_write = wbc->nr_to_write;
824 desired = BIO_MAX_PAGES;
825 if (type == NODE)
826 desired <<= 1;
827
828 wbc->nr_to_write = desired;
829 return desired - nr_to_write;
830}
831
832static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
833{
834 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
835 bool wakeup = false;
836 int i;
837
838 if (force)
839 goto wake_up;
840
841 mutex_lock(&dcc->cmd_lock);
842 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
843 if (i + 1 < dcc->discard_granularity)
844 break;
845 if (!list_empty(&dcc->pend_list[i])) {
846 wakeup = true;
847 break;
848 }
849 }
850 mutex_unlock(&dcc->cmd_lock);
851 if (!wakeup)
852 return;
853wake_up:
854 dcc->discard_wake = 1;
855 wake_up_interruptible_all(&dcc->discard_wait_queue);
856}
1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * fs/f2fs/segment.h
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
7 */
8#include <linux/blkdev.h>
9#include <linux/backing-dev.h>
10
11/* constant macro */
12#define NULL_SEGNO ((unsigned int)(~0))
13#define NULL_SECNO ((unsigned int)(~0))
14
15#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
16#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
17
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 */
20
21#define INVALID_MTIME ULLONG_MAX /* no valid blocks in a segment/section */
22
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)
26
27#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
28#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
29#define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA))
30
31static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
32 unsigned short seg_type)
33{
34 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
35}
36
37#define IS_CURSEG(sbi, seg) \
38 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
39 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
40 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
41 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
42 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
43 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \
44 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \
45 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
46
47#define IS_CURSEC(sbi, secno) \
48 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
49 SEGS_PER_SEC(sbi)) || \
50 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
51 SEGS_PER_SEC(sbi)) || \
52 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
53 SEGS_PER_SEC(sbi)) || \
54 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
55 SEGS_PER_SEC(sbi)) || \
56 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
57 SEGS_PER_SEC(sbi)) || \
58 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
59 SEGS_PER_SEC(sbi)) || \
60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
61 SEGS_PER_SEC(sbi)) || \
62 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
63 SEGS_PER_SEC(sbi)))
64
65#define MAIN_BLKADDR(sbi) \
66 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
67 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
68#define SEG0_BLKADDR(sbi) \
69 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
70 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
71
72#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
73#define MAIN_SECS(sbi) ((sbi)->total_sections)
74
75#define TOTAL_SEGS(sbi) \
76 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
77 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
78#define TOTAL_BLKS(sbi) (SEGS_TO_BLKS(sbi, TOTAL_SEGS(sbi)))
79
80#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
81#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
82 (sbi)->log_blocks_per_seg))
83
84#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
85 (SEGS_TO_BLKS(sbi, GET_R2L_SEGNO(FREE_I(sbi), segno))))
86
87#define NEXT_FREE_BLKADDR(sbi, curseg) \
88 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
89
90#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
91#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
92 (BLKS_TO_SEGS(sbi, GET_SEGOFF_FROM_SEG0(sbi, blk_addr)))
93#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
94 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (BLKS_PER_SEG(sbi) - 1))
95
96#define GET_SEGNO(sbi, blk_addr) \
97 ((!__is_valid_data_blkaddr(blk_addr)) ? \
98 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
99 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
100#define CAP_BLKS_PER_SEC(sbi) \
101 (BLKS_PER_SEC(sbi) - (sbi)->unusable_blocks_per_sec)
102#define CAP_SEGS_PER_SEC(sbi) \
103 (SEGS_PER_SEC(sbi) - \
104 BLKS_TO_SEGS(sbi, (sbi)->unusable_blocks_per_sec))
105#define GET_SEC_FROM_SEG(sbi, segno) \
106 (((segno) == -1) ? -1 : (segno) / SEGS_PER_SEC(sbi))
107#define GET_SEG_FROM_SEC(sbi, secno) \
108 ((secno) * SEGS_PER_SEC(sbi))
109#define GET_ZONE_FROM_SEC(sbi, secno) \
110 (((secno) == -1) ? -1 : (secno) / (sbi)->secs_per_zone)
111#define GET_ZONE_FROM_SEG(sbi, segno) \
112 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
113
114#define GET_SUM_BLOCK(sbi, segno) \
115 ((sbi)->sm_info->ssa_blkaddr + (segno))
116
117#define GET_SUM_TYPE(footer) ((footer)->entry_type)
118#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
119
120#define SIT_ENTRY_OFFSET(sit_i, segno) \
121 ((segno) % (sit_i)->sents_per_block)
122#define SIT_BLOCK_OFFSET(segno) \
123 ((segno) / SIT_ENTRY_PER_BLOCK)
124#define START_SEGNO(segno) \
125 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
126#define SIT_BLK_CNT(sbi) \
127 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
128#define f2fs_bitmap_size(nr) \
129 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
130
131#define SECTOR_FROM_BLOCK(blk_addr) \
132 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
133#define SECTOR_TO_BLOCK(sectors) \
134 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
135
136/*
137 * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
138 * LFS writes data sequentially with cleaning operations.
139 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
140 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
141 * fragmented segment which has similar aging degree.
142 */
143enum {
144 LFS = 0,
145 SSR,
146 AT_SSR,
147};
148
149/*
150 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
151 * GC_CB is based on cost-benefit algorithm.
152 * GC_GREEDY is based on greedy algorithm.
153 * GC_AT is based on age-threshold algorithm.
154 */
155enum {
156 GC_CB = 0,
157 GC_GREEDY,
158 GC_AT,
159 ALLOC_NEXT,
160 FLUSH_DEVICE,
161 MAX_GC_POLICY,
162};
163
164/*
165 * BG_GC means the background cleaning job.
166 * FG_GC means the on-demand cleaning job.
167 */
168enum {
169 BG_GC = 0,
170 FG_GC,
171};
172
173/* for a function parameter to select a victim segment */
174struct victim_sel_policy {
175 int alloc_mode; /* LFS or SSR */
176 int gc_mode; /* GC_CB or GC_GREEDY */
177 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
178 unsigned int max_search; /*
179 * maximum # of segments/sections
180 * to search
181 */
182 unsigned int offset; /* last scanned bitmap offset */
183 unsigned int ofs_unit; /* bitmap search unit */
184 unsigned int min_cost; /* minimum cost */
185 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
186 unsigned int min_segno; /* segment # having min. cost */
187 unsigned long long age; /* mtime of GCed section*/
188 unsigned long long age_threshold;/* age threshold */
189 bool one_time_gc; /* one time GC */
190};
191
192struct seg_entry {
193 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
194 unsigned int valid_blocks:10; /* # of valid blocks */
195 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
196 unsigned int padding:6; /* padding */
197 unsigned char *cur_valid_map; /* validity bitmap of blocks */
198#ifdef CONFIG_F2FS_CHECK_FS
199 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
200#endif
201 /*
202 * # of valid blocks and the validity bitmap stored in the last
203 * checkpoint pack. This information is used by the SSR mode.
204 */
205 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
206 unsigned char *discard_map;
207 unsigned long long mtime; /* modification time of the segment */
208};
209
210struct sec_entry {
211 unsigned int valid_blocks; /* # of valid blocks in a section */
212};
213
214#define MAX_SKIP_GC_COUNT 16
215
216struct revoke_entry {
217 struct list_head list;
218 block_t old_addr; /* for revoking when fail to commit */
219 pgoff_t index;
220};
221
222struct sit_info {
223 block_t sit_base_addr; /* start block address of SIT area */
224 block_t sit_blocks; /* # of blocks used by SIT area */
225 block_t written_valid_blocks; /* # of valid blocks in main area */
226 char *bitmap; /* all bitmaps pointer */
227 char *sit_bitmap; /* SIT bitmap pointer */
228#ifdef CONFIG_F2FS_CHECK_FS
229 char *sit_bitmap_mir; /* SIT bitmap mirror */
230
231 /* bitmap of segments to be ignored by GC in case of errors */
232 unsigned long *invalid_segmap;
233#endif
234 unsigned int bitmap_size; /* SIT bitmap size */
235
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 */
243
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 */
249 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
250 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
251
252 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
253};
254
255struct free_segmap_info {
256 unsigned int start_segno; /* start segment number logically */
257 unsigned int free_segments; /* # of free segments */
258 unsigned int free_sections; /* # of free sections */
259 spinlock_t segmap_lock; /* free segmap lock */
260 unsigned long *free_segmap; /* free segment bitmap */
261 unsigned long *free_secmap; /* free section bitmap */
262};
263
264/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
265enum dirty_type {
266 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
267 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
268 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
269 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
270 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
271 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
272 DIRTY, /* to count # of dirty segments */
273 PRE, /* to count # of entirely obsolete segments */
274 NR_DIRTY_TYPE
275};
276
277struct dirty_seglist_info {
278 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
279 unsigned long *dirty_secmap;
280 struct mutex seglist_lock; /* lock for segment bitmaps */
281 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
282 unsigned long *victim_secmap; /* background GC victims */
283 unsigned long *pinned_secmap; /* pinned victims from foreground GC */
284 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */
285 bool enable_pin_section; /* enable pinning section */
286};
287
288/* for active log information */
289struct curseg_info {
290 struct mutex curseg_mutex; /* lock for consistency */
291 struct f2fs_summary_block *sum_blk; /* cached summary block */
292 struct rw_semaphore journal_rwsem; /* protect journal area */
293 struct f2fs_journal *journal; /* cached journal info */
294 unsigned char alloc_type; /* current allocation type */
295 unsigned short seg_type; /* segment type like CURSEG_XXX_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 */
300 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */
301 bool inited; /* indicate inmem log is inited */
302};
303
304struct sit_entry_set {
305 struct list_head set_list; /* link with all sit sets */
306 unsigned int start_segno; /* start segno of sits in set */
307 unsigned int entry_cnt; /* the # of sit entries in set */
308};
309
310/*
311 * inline functions
312 */
313static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
314{
315 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
316}
317
318static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
319 unsigned int segno)
320{
321 struct sit_info *sit_i = SIT_I(sbi);
322 return &sit_i->sentries[segno];
323}
324
325static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
326 unsigned int segno)
327{
328 struct sit_info *sit_i = SIT_I(sbi);
329 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
330}
331
332static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
333 unsigned int segno, bool use_section)
334{
335 /*
336 * In order to get # of valid blocks in a section instantly from many
337 * segments, f2fs manages two counting structures separately.
338 */
339 if (use_section && __is_large_section(sbi))
340 return get_sec_entry(sbi, segno)->valid_blocks;
341 else
342 return get_seg_entry(sbi, segno)->valid_blocks;
343}
344
345static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
346 unsigned int segno, bool use_section)
347{
348 if (use_section && __is_large_section(sbi)) {
349 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
350 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
351 unsigned int blocks = 0;
352 int i;
353
354 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) {
355 struct seg_entry *se = get_seg_entry(sbi, start_segno);
356
357 blocks += se->ckpt_valid_blocks;
358 }
359 return blocks;
360 }
361 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
362}
363
364static inline void seg_info_from_raw_sit(struct seg_entry *se,
365 struct f2fs_sit_entry *rs)
366{
367 se->valid_blocks = GET_SIT_VBLOCKS(rs);
368 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
369 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
370 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
371#ifdef CONFIG_F2FS_CHECK_FS
372 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
373#endif
374 se->type = GET_SIT_TYPE(rs);
375 se->mtime = le64_to_cpu(rs->mtime);
376}
377
378static inline void __seg_info_to_raw_sit(struct seg_entry *se,
379 struct f2fs_sit_entry *rs)
380{
381 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
382 se->valid_blocks;
383 rs->vblocks = cpu_to_le16(raw_vblocks);
384 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
385 rs->mtime = cpu_to_le64(se->mtime);
386}
387
388static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
389 struct page *page, unsigned int start)
390{
391 struct f2fs_sit_block *raw_sit;
392 struct seg_entry *se;
393 struct f2fs_sit_entry *rs;
394 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
395 (unsigned long)MAIN_SEGS(sbi));
396 int i;
397
398 raw_sit = (struct f2fs_sit_block *)page_address(page);
399 memset(raw_sit, 0, PAGE_SIZE);
400 for (i = 0; i < end - start; i++) {
401 rs = &raw_sit->entries[i];
402 se = get_seg_entry(sbi, start + i);
403 __seg_info_to_raw_sit(se, rs);
404 }
405}
406
407static inline void seg_info_to_raw_sit(struct seg_entry *se,
408 struct f2fs_sit_entry *rs)
409{
410 __seg_info_to_raw_sit(se, rs);
411
412 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
413 se->ckpt_valid_blocks = se->valid_blocks;
414}
415
416static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
417 unsigned int max, unsigned int segno)
418{
419 unsigned int ret;
420 spin_lock(&free_i->segmap_lock);
421 ret = find_next_bit(free_i->free_segmap, max, segno);
422 spin_unlock(&free_i->segmap_lock);
423 return ret;
424}
425
426static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
427{
428 struct free_segmap_info *free_i = FREE_I(sbi);
429 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
430 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
431 unsigned int next;
432 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi);
433
434 spin_lock(&free_i->segmap_lock);
435 clear_bit(segno, free_i->free_segmap);
436 free_i->free_segments++;
437
438 next = find_next_bit(free_i->free_segmap,
439 start_segno + SEGS_PER_SEC(sbi), start_segno);
440 if (next >= start_segno + usable_segs) {
441 clear_bit(secno, free_i->free_secmap);
442 free_i->free_sections++;
443 }
444 spin_unlock(&free_i->segmap_lock);
445}
446
447static inline void __set_inuse(struct f2fs_sb_info *sbi,
448 unsigned int segno)
449{
450 struct free_segmap_info *free_i = FREE_I(sbi);
451 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
452
453 set_bit(segno, free_i->free_segmap);
454 free_i->free_segments--;
455 if (!test_and_set_bit(secno, free_i->free_secmap))
456 free_i->free_sections--;
457}
458
459static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
460 unsigned int segno, bool inmem)
461{
462 struct free_segmap_info *free_i = FREE_I(sbi);
463 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
464 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
465 unsigned int next;
466 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi);
467
468 spin_lock(&free_i->segmap_lock);
469 if (test_and_clear_bit(segno, free_i->free_segmap)) {
470 free_i->free_segments++;
471
472 if (!inmem && IS_CURSEC(sbi, secno))
473 goto skip_free;
474 next = find_next_bit(free_i->free_segmap,
475 start_segno + SEGS_PER_SEC(sbi), start_segno);
476 if (next >= start_segno + usable_segs) {
477 if (test_and_clear_bit(secno, free_i->free_secmap))
478 free_i->free_sections++;
479 }
480 }
481skip_free:
482 spin_unlock(&free_i->segmap_lock);
483}
484
485static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
486 unsigned int segno)
487{
488 struct free_segmap_info *free_i = FREE_I(sbi);
489 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
490
491 spin_lock(&free_i->segmap_lock);
492 if (!test_and_set_bit(segno, free_i->free_segmap)) {
493 free_i->free_segments--;
494 if (!test_and_set_bit(secno, free_i->free_secmap))
495 free_i->free_sections--;
496 }
497 spin_unlock(&free_i->segmap_lock);
498}
499
500static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
501 void *dst_addr)
502{
503 struct sit_info *sit_i = SIT_I(sbi);
504
505#ifdef CONFIG_F2FS_CHECK_FS
506 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
507 sit_i->bitmap_size))
508 f2fs_bug_on(sbi, 1);
509#endif
510 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
511}
512
513static inline block_t written_block_count(struct f2fs_sb_info *sbi)
514{
515 return SIT_I(sbi)->written_valid_blocks;
516}
517
518static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
519{
520 return FREE_I(sbi)->free_segments;
521}
522
523static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
524{
525 return SM_I(sbi)->reserved_segments;
526}
527
528static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
529{
530 return FREE_I(sbi)->free_sections;
531}
532
533static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
534{
535 return DIRTY_I(sbi)->nr_dirty[PRE];
536}
537
538static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
539{
540 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
541 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
542 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
543 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
544 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
545 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
546}
547
548static inline int overprovision_segments(struct f2fs_sb_info *sbi)
549{
550 return SM_I(sbi)->ovp_segments;
551}
552
553static inline int reserved_sections(struct f2fs_sb_info *sbi)
554{
555 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
556}
557
558static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
559 unsigned int node_blocks, unsigned int data_blocks,
560 unsigned int dent_blocks)
561{
562
563 unsigned int segno, left_blocks, blocks;
564 int i;
565
566 /* check current data/node sections in the worst case. */
567 for (i = CURSEG_HOT_DATA; i < NR_PERSISTENT_LOG; i++) {
568 segno = CURSEG_I(sbi, i)->segno;
569 left_blocks = CAP_BLKS_PER_SEC(sbi) -
570 get_ckpt_valid_blocks(sbi, segno, true);
571
572 blocks = i <= CURSEG_COLD_DATA ? data_blocks : node_blocks;
573 if (blocks > left_blocks)
574 return false;
575 }
576
577 /* check current data section for dentry blocks. */
578 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
579 left_blocks = CAP_BLKS_PER_SEC(sbi) -
580 get_ckpt_valid_blocks(sbi, segno, true);
581 if (dent_blocks > left_blocks)
582 return false;
583 return true;
584}
585
586/*
587 * calculate needed sections for dirty node/dentry and call
588 * has_curseg_enough_space, please note that, it needs to account
589 * dirty data as well in lfs mode when checkpoint is disabled.
590 */
591static inline void __get_secs_required(struct f2fs_sb_info *sbi,
592 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p)
593{
594 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
595 get_pages(sbi, F2FS_DIRTY_DENTS) +
596 get_pages(sbi, F2FS_DIRTY_IMETA);
597 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
598 unsigned int total_data_blocks = 0;
599 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
600 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
601 unsigned int data_secs = 0;
602 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
603 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
604 unsigned int data_blocks = 0;
605
606 if (f2fs_lfs_mode(sbi) &&
607 unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
608 total_data_blocks = get_pages(sbi, F2FS_DIRTY_DATA);
609 data_secs = total_data_blocks / CAP_BLKS_PER_SEC(sbi);
610 data_blocks = total_data_blocks % CAP_BLKS_PER_SEC(sbi);
611 }
612
613 if (lower_p)
614 *lower_p = node_secs + dent_secs + data_secs;
615 if (upper_p)
616 *upper_p = node_secs + dent_secs +
617 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0) +
618 (data_blocks ? 1 : 0);
619 if (curseg_p)
620 *curseg_p = has_curseg_enough_space(sbi,
621 node_blocks, data_blocks, dent_blocks);
622}
623
624static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
625 int freed, int needed)
626{
627 unsigned int free_secs, lower_secs, upper_secs;
628 bool curseg_space;
629
630 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
631 return false;
632
633 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space);
634
635 free_secs = free_sections(sbi) + freed;
636 lower_secs += needed + reserved_sections(sbi);
637 upper_secs += needed + reserved_sections(sbi);
638
639 if (free_secs > upper_secs)
640 return false;
641 if (free_secs <= lower_secs)
642 return true;
643 return !curseg_space;
644}
645
646static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi,
647 int freed, int needed)
648{
649 return !has_not_enough_free_secs(sbi, freed, needed);
650}
651
652static inline bool has_enough_free_blks(struct f2fs_sb_info *sbi)
653{
654 unsigned int total_free_blocks = 0;
655 unsigned int avail_user_block_count;
656
657 spin_lock(&sbi->stat_lock);
658
659 avail_user_block_count = get_available_block_count(sbi, NULL, true);
660 total_free_blocks = avail_user_block_count - (unsigned int)valid_user_blocks(sbi);
661
662 spin_unlock(&sbi->stat_lock);
663
664 return total_free_blocks > 0;
665}
666
667static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
668{
669 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
670 return true;
671 if (likely(has_enough_free_secs(sbi, 0, 0)))
672 return true;
673 if (!f2fs_lfs_mode(sbi) &&
674 likely(has_enough_free_blks(sbi)))
675 return true;
676 return false;
677}
678
679static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
680{
681 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
682}
683
684static inline int utilization(struct f2fs_sb_info *sbi)
685{
686 return div_u64((u64)valid_user_blocks(sbi) * 100,
687 sbi->user_block_count);
688}
689
690/*
691 * Sometimes f2fs may be better to drop out-of-place update policy.
692 * And, users can control the policy through sysfs entries.
693 * There are five policies with triggering conditions as follows.
694 * F2FS_IPU_FORCE - all the time,
695 * F2FS_IPU_SSR - if SSR mode is activated,
696 * F2FS_IPU_UTIL - if FS utilization is over threashold,
697 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
698 * threashold,
699 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
700 * storages. IPU will be triggered only if the # of dirty
701 * pages over min_fsync_blocks. (=default option)
702 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
703 * F2FS_IPU_NOCACHE - disable IPU bio cache.
704 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
705 * FI_OPU_WRITE flag.
706 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
707 */
708#define DEF_MIN_IPU_UTIL 70
709#define DEF_MIN_FSYNC_BLOCKS 8
710#define DEF_MIN_HOT_BLOCKS 16
711
712#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
713
714#define F2FS_IPU_DISABLE 0
715
716/* Modification on enum should be synchronized with ipu_mode_names array */
717enum {
718 F2FS_IPU_FORCE,
719 F2FS_IPU_SSR,
720 F2FS_IPU_UTIL,
721 F2FS_IPU_SSR_UTIL,
722 F2FS_IPU_FSYNC,
723 F2FS_IPU_ASYNC,
724 F2FS_IPU_NOCACHE,
725 F2FS_IPU_HONOR_OPU_WRITE,
726 F2FS_IPU_MAX,
727};
728
729static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi)
730{
731 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE;
732}
733
734#define F2FS_IPU_POLICY(name) \
735static inline bool IS_##name(struct f2fs_sb_info *sbi) \
736{ \
737 return SM_I(sbi)->ipu_policy & BIT(name); \
738}
739
740F2FS_IPU_POLICY(F2FS_IPU_FORCE);
741F2FS_IPU_POLICY(F2FS_IPU_SSR);
742F2FS_IPU_POLICY(F2FS_IPU_UTIL);
743F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL);
744F2FS_IPU_POLICY(F2FS_IPU_FSYNC);
745F2FS_IPU_POLICY(F2FS_IPU_ASYNC);
746F2FS_IPU_POLICY(F2FS_IPU_NOCACHE);
747F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE);
748
749static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
750 int type)
751{
752 struct curseg_info *curseg = CURSEG_I(sbi, type);
753 return curseg->segno;
754}
755
756static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
757 int type)
758{
759 struct curseg_info *curseg = CURSEG_I(sbi, type);
760 return curseg->alloc_type;
761}
762
763static inline bool valid_main_segno(struct f2fs_sb_info *sbi,
764 unsigned int segno)
765{
766 return segno <= (MAIN_SEGS(sbi) - 1);
767}
768
769static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
770{
771 struct f2fs_sb_info *sbi = fio->sbi;
772
773 if (__is_valid_data_blkaddr(fio->old_blkaddr))
774 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
775 META_GENERIC : DATA_GENERIC);
776 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
777 META_GENERIC : DATA_GENERIC_ENHANCE);
778}
779
780/*
781 * Summary block is always treated as an invalid block
782 */
783static inline int check_block_count(struct f2fs_sb_info *sbi,
784 int segno, struct f2fs_sit_entry *raw_sit)
785{
786 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
787 int valid_blocks = 0;
788 int cur_pos = 0, next_pos;
789 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
790
791 /* check bitmap with valid block count */
792 do {
793 if (is_valid) {
794 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
795 usable_blks_per_seg,
796 cur_pos);
797 valid_blocks += next_pos - cur_pos;
798 } else
799 next_pos = find_next_bit_le(&raw_sit->valid_map,
800 usable_blks_per_seg,
801 cur_pos);
802 cur_pos = next_pos;
803 is_valid = !is_valid;
804 } while (cur_pos < usable_blks_per_seg);
805
806 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
807 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
808 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
809 set_sbi_flag(sbi, SBI_NEED_FSCK);
810 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
811 return -EFSCORRUPTED;
812 }
813
814 if (usable_blks_per_seg < BLKS_PER_SEG(sbi))
815 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
816 BLKS_PER_SEG(sbi),
817 usable_blks_per_seg) != BLKS_PER_SEG(sbi));
818
819 /* check segment usage, and check boundary of a given segment number */
820 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
821 || !valid_main_segno(sbi, segno))) {
822 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
823 GET_SIT_VBLOCKS(raw_sit), segno);
824 set_sbi_flag(sbi, SBI_NEED_FSCK);
825 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
826 return -EFSCORRUPTED;
827 }
828 return 0;
829}
830
831static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
832 unsigned int start)
833{
834 struct sit_info *sit_i = SIT_I(sbi);
835 unsigned int offset = SIT_BLOCK_OFFSET(start);
836 block_t blk_addr = sit_i->sit_base_addr + offset;
837
838 f2fs_bug_on(sbi, !valid_main_segno(sbi, start));
839
840#ifdef CONFIG_F2FS_CHECK_FS
841 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
842 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
843 f2fs_bug_on(sbi, 1);
844#endif
845
846 /* calculate sit block address */
847 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
848 blk_addr += sit_i->sit_blocks;
849
850 return blk_addr;
851}
852
853static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
854 pgoff_t block_addr)
855{
856 struct sit_info *sit_i = SIT_I(sbi);
857 block_addr -= sit_i->sit_base_addr;
858 if (block_addr < sit_i->sit_blocks)
859 block_addr += sit_i->sit_blocks;
860 else
861 block_addr -= sit_i->sit_blocks;
862
863 return block_addr + sit_i->sit_base_addr;
864}
865
866static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
867{
868 unsigned int block_off = SIT_BLOCK_OFFSET(start);
869
870 f2fs_change_bit(block_off, sit_i->sit_bitmap);
871#ifdef CONFIG_F2FS_CHECK_FS
872 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
873#endif
874}
875
876static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
877 bool base_time)
878{
879 struct sit_info *sit_i = SIT_I(sbi);
880 time64_t diff, now = ktime_get_boottime_seconds();
881
882 if (now >= sit_i->mounted_time)
883 return sit_i->elapsed_time + now - sit_i->mounted_time;
884
885 /* system time is set to the past */
886 if (!base_time) {
887 diff = sit_i->mounted_time - now;
888 if (sit_i->elapsed_time >= diff)
889 return sit_i->elapsed_time - diff;
890 return 0;
891 }
892 return sit_i->elapsed_time;
893}
894
895static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
896 unsigned int ofs_in_node, unsigned char version)
897{
898 sum->nid = cpu_to_le32(nid);
899 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
900 sum->version = version;
901}
902
903static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
904{
905 return __start_cp_addr(sbi) +
906 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
907}
908
909static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
910{
911 return __start_cp_addr(sbi) +
912 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
913 - (base + 1) + type;
914}
915
916static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
917{
918 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
919 return true;
920 return false;
921}
922
923/*
924 * It is very important to gather dirty pages and write at once, so that we can
925 * submit a big bio without interfering other data writes.
926 * By default, 512 pages for directory data,
927 * 512 pages (2MB) * 8 for nodes, and
928 * 256 pages * 8 for meta are set.
929 */
930static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
931{
932 if (sbi->sb->s_bdi->wb.dirty_exceeded)
933 return 0;
934
935 if (type == DATA)
936 return BLKS_PER_SEG(sbi);
937 else if (type == NODE)
938 return SEGS_TO_BLKS(sbi, 8);
939 else if (type == META)
940 return 8 * BIO_MAX_VECS;
941 else
942 return 0;
943}
944
945/*
946 * When writing pages, it'd better align nr_to_write for segment size.
947 */
948static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
949 struct writeback_control *wbc)
950{
951 long nr_to_write, desired;
952
953 if (wbc->sync_mode != WB_SYNC_NONE)
954 return 0;
955
956 nr_to_write = wbc->nr_to_write;
957 desired = BIO_MAX_VECS;
958 if (type == NODE)
959 desired <<= 1;
960
961 wbc->nr_to_write = desired;
962 return desired - nr_to_write;
963}
964
965static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
966{
967 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
968 bool wakeup = false;
969 int i;
970
971 if (force)
972 goto wake_up;
973
974 mutex_lock(&dcc->cmd_lock);
975 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
976 if (i + 1 < dcc->discard_granularity)
977 break;
978 if (!list_empty(&dcc->pend_list[i])) {
979 wakeup = true;
980 break;
981 }
982 }
983 mutex_unlock(&dcc->cmd_lock);
984 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
985 return;
986wake_up:
987 dcc->discard_wake = true;
988 wake_up_interruptible_all(&dcc->discard_wait_queue);
989}