projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'vfio-v4.20-rc1.v2' of git://github.com/awilliam/linux-vfio
[linux-2.6-block.git] / fs / f2fs / segment.h
CommitLineData
7c1a000d 1// SPDX-License-Identifier: GPL-2.0
0a8165d7 2/*
39a53e0c
JK		 3 * fs/f2fs/segment.h
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
39a53e0c 7 */
ac5d156c 8#include <linux/blkdev.h>
66114cad 9#include <linux/backing-dev.h>
ac5d156c 10
39a53e0c
JK		 11/* constant macro */
12#define NULL_SEGNO ((unsigned int)(~0))
5ec4e49f 13#define NULL_SECNO ((unsigned int)(~0))
39a53e0c 14
58c41035 15#define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
44a83499 16#define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
81eb8d6e 17
2040fce8
JK		 18#define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
19
6224da87 20/* L: Logical segment # in volume, R: Relative segment # in main area */
68afcf2d
TK		 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)
39a53e0c 23
68afcf2d
TK		 24#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
25#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE)
39a53e0c 26
a912b54d
JK		 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)
30
5c773ba3 31#define IS_CURSEG(sbi, seg) \
68afcf2d
TK		 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))
39a53e0c
JK		 38
39#define IS_CURSEC(sbi, secno) \
68afcf2d
TK		 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)) \
39a53e0c 52
0833721e
YH		 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))
7cd8558b
JK		 59
60#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
68afcf2d 61#define MAIN_SECS(sbi) ((sbi)->total_sections)
7cd8558b 62
0833721e
YH		 63#define TOTAL_SEGS(sbi) \
64 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
65 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
68afcf2d 66#define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
7cd8558b
JK		 67
68#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
68afcf2d
TK		 69#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
70 (sbi)->log_blocks_per_seg))
7cd8558b
JK		 71
72#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
68afcf2d 73 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
7cd8558b 74
39a53e0c 75#define NEXT_FREE_BLKADDR(sbi, curseg) \
68afcf2d 76 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
39a53e0c 77
7cd8558b 78#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
39a53e0c 79#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
68afcf2d 80 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
491c0854 81#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
68afcf2d 82 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
491c0854 83
39a53e0c 84#define GET_SEGNO(sbi, blk_addr) \
e1da7872 85 ((!is_valid_data_blkaddr(sbi, blk_addr)) ? \
39a53e0c
JK		 86 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
87 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
4ddb1a4d
JK		 88#define BLKS_PER_SEC(sbi) \
89 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
90#define GET_SEC_FROM_SEG(sbi, segno) \
68afcf2d 91 ((segno) / (sbi)->segs_per_sec)
4ddb1a4d 92#define GET_SEG_FROM_SEC(sbi, secno) \
63fcf8e8 93 ((secno) * (sbi)->segs_per_sec)
4ddb1a4d
JK		 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))
39a53e0c
JK		 98
99#define GET_SUM_BLOCK(sbi, segno) \
68afcf2d 100 ((sbi)->sm_info->ssa_blkaddr + (segno))
39a53e0c
JK		 101
102#define GET_SUM_TYPE(footer) ((footer)->entry_type)
68afcf2d 103#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
39a53e0c
JK		 104
105#define SIT_ENTRY_OFFSET(sit_i, segno) \
68afcf2d 106 ((segno) % (sit_i)->sents_per_block)
d3a14afd 107#define SIT_BLOCK_OFFSET(segno) \
68afcf2d 108 ((segno) / SIT_ENTRY_PER_BLOCK)
d3a14afd
CY		 109#define START_SEGNO(segno) \
110 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
74de593a 111#define SIT_BLK_CNT(sbi) \
7cd8558b 112 ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
39a53e0c
JK		 113#define f2fs_bitmap_size(nr) \
114 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
39a53e0c 115
55cf9cb6
CY		 116#define SECTOR_FROM_BLOCK(blk_addr) \
117 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
118#define SECTOR_TO_BLOCK(sectors) \
68afcf2d 119 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
3cd8a239 120
39a53e0c
JK		 121/*
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.
125 */
126enum {
127 ALLOC_RIGHT = 0,
128 ALLOC_LEFT
129};
130
131/*
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.
135 */
136enum {
137 LFS = 0,
138 SSR
139};
140
141/*
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.
145 */
146enum {
147 GC_CB = 0,
e066b83c
JK		 148 GC_GREEDY,
149 ALLOC_NEXT,
150 FLUSH_DEVICE,
151 MAX_GC_POLICY,
39a53e0c
JK		 152};
153
154/*
155 * BG_GC means the background cleaning job.
156 * FG_GC means the on-demand cleaning job.
6aefd93b 157 * FORCE_FG_GC means on-demand cleaning job in background.
39a53e0c
JK		 158 */
159enum {
160 BG_GC = 0,
6aefd93b
JK		 161 FG_GC,
162 FORCE_FG_GC,
39a53e0c
JK		 163};
164
165/* for a function parameter to select a victim segment */
166struct 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 */
a26b7c8a 170 unsigned int max_search; /* maximum # of segments to search */
39a53e0c
JK		 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 */
175};
176
177struct seg_entry {
f51b4ce6
CY		 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 */
39a53e0c 182 unsigned char *cur_valid_map; /* validity bitmap of blocks */
355e7891
CY		 183#ifdef CONFIG_F2FS_CHECK_FS
184 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
185#endif
39a53e0c
JK		 186 /*
187 * # of valid blocks and the validity bitmap stored in the the last
188 * checkpoint pack. This information is used by the SSR mode.
189 */
f51b4ce6 190 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
a66cdd98 191 unsigned char *discard_map;
39a53e0c
JK		 192 unsigned long long mtime; /* modification time of the segment */
193};
194
195struct sec_entry {
196 unsigned int valid_blocks; /* # of valid blocks in a section */
197};
198
199struct segment_allocation {
200 void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
201};
202
decd36b6
CY		 203/*
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.
206 */
d48dfc20 207#define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
0a595eba 208#define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
decd36b6
CY		 209
210#define IS_ATOMIC_WRITTEN_PAGE(page) \
211 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
0a595eba
JK		 212#define IS_DUMMY_WRITTEN_PAGE(page) \
213 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
decd36b6 214
6f8d4455 215#define MAX_SKIP_GC_COUNT 16
2ef79ecb 216
88b88a66
JK		 217struct inmem_pages {
218 struct list_head list;
219 struct page *page;
28bc106b 220 block_t old_addr; /* for revoking when fail to commit */
88b88a66
JK		 221};
222
39a53e0c
JK		 223struct sit_info {
224 const struct segment_allocation *s_ops;
225
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 *sit_bitmap; /* SIT bitmap pointer */
ae27d62e
CY		 230#ifdef CONFIG_F2FS_CHECK_FS
231 char *sit_bitmap_mir; /* SIT bitmap mirror */
232#endif
39a53e0c
JK		 233 unsigned int bitmap_size; /* SIT bitmap size */
234
60a3b782 235 unsigned long *tmp_map; /* bitmap for temporal use */
39a53e0c
JK		 236 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
237 unsigned int dirty_sentries; /* # of dirty sentries */
238 unsigned int sents_per_block; /* # of SIT entries per block */
3d26fa6b 239 struct rw_semaphore sentry_lock; /* to protect SIT cache */
39a53e0c
JK		 240 struct seg_entry *sentries; /* SIT segment-level cache */
241 struct sec_entry *sec_entries; /* SIT section-level cache */
242
243 /* for cost-benefit algorithm in cleaning procedure */
244 unsigned long long elapsed_time; /* elapsed time after mount */
245 unsigned long long mounted_time; /* mount time */
246 unsigned long long min_mtime; /* min. modification time */
247 unsigned long long max_mtime; /* max. modification time */
e066b83c
JK		 248
249 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
39a53e0c
JK		 250};
251
252struct free_segmap_info {
253 unsigned int start_segno; /* start segment number logically */
254 unsigned int free_segments; /* # of free segments */
255 unsigned int free_sections; /* # of free sections */
1a118ccf 256 spinlock_t segmap_lock; /* free segmap lock */
39a53e0c
JK		 257 unsigned long *free_segmap; /* free segment bitmap */
258 unsigned long *free_secmap; /* free section bitmap */
259};
260
261/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
262enum dirty_type {
263 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
264 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
265 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
266 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
267 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
268 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
269 DIRTY, /* to count # of dirty segments */
270 PRE, /* to count # of entirely obsolete segments */
271 NR_DIRTY_TYPE
272};
273
274struct dirty_seglist_info {
275 const struct victim_selection *v_ops; /* victim selction operation */
276 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
277 struct mutex seglist_lock; /* lock for segment bitmaps */
278 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
5ec4e49f 279 unsigned long *victim_secmap; /* background GC victims */
39a53e0c
JK		 280};
281
282/* victim selection function for cleaning and SSR */
283struct victim_selection {
284 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
285 int, int, char);
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 */
b7ad7512
CY		 292 struct rw_semaphore journal_rwsem; /* protect journal area */
293 struct f2fs_journal *journal; /* cached journal info */
39a53e0c
JK		 294 unsigned char alloc_type; /* current allocation type */
295 unsigned int segno; /* current segment number */
296 unsigned short next_blkoff; /* next block offset to write */
297 unsigned int zone; /* current zone number */
298 unsigned int next_segno; /* preallocated segment */
299};
300
184a5cd2
CY		 301struct sit_entry_set {
302 struct list_head set_list; /* link with all sit sets */
303 unsigned int start_segno; /* start segno of sits in set */
304 unsigned int entry_cnt; /* the # of sit entries in set */
305};
306
39a53e0c
JK		 307/*
308 * inline functions
309 */
310static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
311{
312 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
313}
314
315static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
316 unsigned int segno)
317{
318 struct sit_info *sit_i = SIT_I(sbi);
319 return &sit_i->sentries[segno];
320}
321
322static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
323 unsigned int segno)
324{
325 struct sit_info *sit_i = SIT_I(sbi);
4ddb1a4d 326 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
39a53e0c
JK		 327}
328
329static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
302bd348 330 unsigned int segno, bool use_section)
39a53e0c
JK		 331{
332 /*
333 * In order to get # of valid blocks in a section instantly from many
334 * segments, f2fs manages two counting structures separately.
335 */
302bd348 336 if (use_section && sbi->segs_per_sec > 1)
39a53e0c
JK		 337 return get_sec_entry(sbi, segno)->valid_blocks;
338 else
339 return get_seg_entry(sbi, segno)->valid_blocks;
340}
341
4354994f
DR		 342static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
343 unsigned int segno)
344{
345 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
346}
347
39a53e0c
JK		 348static inline void seg_info_from_raw_sit(struct seg_entry *se,
349 struct f2fs_sit_entry *rs)
350{
351 se->valid_blocks = GET_SIT_VBLOCKS(rs);
352 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
353 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
354 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
355e7891
CY		 355#ifdef CONFIG_F2FS_CHECK_FS
356 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
357#endif
39a53e0c
JK		 358 se->type = GET_SIT_TYPE(rs);
359 se->mtime = le64_to_cpu(rs->mtime);
360}
361
068c3cd8 362static inline void __seg_info_to_raw_sit(struct seg_entry *se,
39a53e0c
JK		 363 struct f2fs_sit_entry *rs)
364{
365 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
366 se->valid_blocks;
367 rs->vblocks = cpu_to_le16(raw_vblocks);
368 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
068c3cd8
YH		 369 rs->mtime = cpu_to_le64(se->mtime);
370}
371
372static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
373 struct page *page, unsigned int start)
374{
375 struct f2fs_sit_block *raw_sit;
376 struct seg_entry *se;
377 struct f2fs_sit_entry *rs;
378 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
379 (unsigned long)MAIN_SEGS(sbi));
380 int i;
381
382 raw_sit = (struct f2fs_sit_block *)page_address(page);
81114baa 383 memset(raw_sit, 0, PAGE_SIZE);
068c3cd8
YH		 384 for (i = 0; i < end - start; i++) {
385 rs = &raw_sit->entries[i];
386 se = get_seg_entry(sbi, start + i);
387 __seg_info_to_raw_sit(se, rs);
388 }
389}
390
391static inline void seg_info_to_raw_sit(struct seg_entry *se,
392 struct f2fs_sit_entry *rs)
393{
394 __seg_info_to_raw_sit(se, rs);
395
39a53e0c
JK		 396 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
397 se->ckpt_valid_blocks = se->valid_blocks;
39a53e0c
JK		 398}
399
400static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
401 unsigned int max, unsigned int segno)
402{
403 unsigned int ret;
1a118ccf 404 spin_lock(&free_i->segmap_lock);
39a53e0c 405 ret = find_next_bit(free_i->free_segmap, max, segno);
1a118ccf 406 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 407 return ret;
408}
409
410static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
411{
412 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 413 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
414 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK		 415 unsigned int next;
416
1a118ccf 417 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 418 clear_bit(segno, free_i->free_segmap);
419 free_i->free_segments++;
420
7fd97019
WL		 421 next = find_next_bit(free_i->free_segmap,
422 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK		 423 if (next >= start_segno + sbi->segs_per_sec) {
424 clear_bit(secno, free_i->free_secmap);
425 free_i->free_sections++;
426 }
1a118ccf 427 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 428}
429
430static inline void __set_inuse(struct f2fs_sb_info *sbi,
431 unsigned int segno)
432{
433 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 434 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
435
39a53e0c
JK		 436 set_bit(segno, free_i->free_segmap);
437 free_i->free_segments--;
438 if (!test_and_set_bit(secno, free_i->free_secmap))
439 free_i->free_sections--;
440}
441
442static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
443 unsigned int segno)
444{
445 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 446 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
447 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
39a53e0c
JK		 448 unsigned int next;
449
1a118ccf 450 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 451 if (test_and_clear_bit(segno, free_i->free_segmap)) {
452 free_i->free_segments++;
453
3611ce99
YS		 454 if (IS_CURSEC(sbi, secno))
455 goto skip_free;
f1121ab0
CY		 456 next = find_next_bit(free_i->free_segmap,
457 start_segno + sbi->segs_per_sec, start_segno);
39a53e0c
JK		 458 if (next >= start_segno + sbi->segs_per_sec) {
459 if (test_and_clear_bit(secno, free_i->free_secmap))
460 free_i->free_sections++;
461 }
462 }
3611ce99 463skip_free:
1a118ccf 464 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 465}
466
467static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
468 unsigned int segno)
469{
470 struct free_segmap_info *free_i = FREE_I(sbi);
4ddb1a4d
JK		 471 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
472
1a118ccf 473 spin_lock(&free_i->segmap_lock);
39a53e0c
JK		 474 if (!test_and_set_bit(segno, free_i->free_segmap)) {
475 free_i->free_segments--;
476 if (!test_and_set_bit(secno, free_i->free_secmap))
477 free_i->free_sections--;
478 }
1a118ccf 479 spin_unlock(&free_i->segmap_lock);
39a53e0c
JK		 480}
481
482static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
483 void *dst_addr)
484{
485 struct sit_info *sit_i = SIT_I(sbi);
ae27d62e
CY		 486
487#ifdef CONFIG_F2FS_CHECK_FS
488 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
489 sit_i->bitmap_size))
490 f2fs_bug_on(sbi, 1);
491#endif
39a53e0c
JK		 492 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
493}
494
495static inline block_t written_block_count(struct f2fs_sb_info *sbi)
496{
8b8343fa 497 return SIT_I(sbi)->written_valid_blocks;
39a53e0c
JK		 498}
499
500static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
501{
8b8343fa 502 return FREE_I(sbi)->free_segments;
39a53e0c
JK		 503}
504
505static inline int reserved_segments(struct f2fs_sb_info *sbi)
506{
507 return SM_I(sbi)->reserved_segments;
508}
509
510static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
511{
8b8343fa 512 return FREE_I(sbi)->free_sections;
39a53e0c
JK		 513}
514
515static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
516{
517 return DIRTY_I(sbi)->nr_dirty[PRE];
518}
519
520static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
521{
522 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
523 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
524 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
525 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
526 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
527 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
528}
529
530static inline int overprovision_segments(struct f2fs_sb_info *sbi)
531{
532 return SM_I(sbi)->ovp_segments;
533}
534
39a53e0c
JK		 535static inline int reserved_sections(struct f2fs_sb_info *sbi)
536{
4ddb1a4d 537 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
39a53e0c
JK		 538}
539
bf34c93d
CY		 540static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi)
541{
542 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
543 get_pages(sbi, F2FS_DIRTY_DENTS);
544 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
545 unsigned int segno, left_blocks;
546 int i;
547
548 /* check current node segment */
549 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
550 segno = CURSEG_I(sbi, i)->segno;
551 left_blocks = sbi->blocks_per_seg -
552 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
553
554 if (node_blocks > left_blocks)
555 return false;
556 }
557
558 /* check current data segment */
559 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
560 left_blocks = sbi->blocks_per_seg -
561 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
562 if (dent_blocks > left_blocks)
563 return false;
564 return true;
565}
566
7f3037a5
JK		 567static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
568 int freed, int needed)
39a53e0c 569{
5ac206cf
NJ		 570 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
571 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
b9610bdf 572 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
0f18b462 573
caf0047e 574 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
029cd28c
JK		 575 return false;
576
bf34c93d
CY		 577 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed &&
578 has_curseg_enough_space(sbi))
579 return false;
7f3037a5 580 return (free_sections(sbi) + freed) <=
b9610bdf
JK		 581 (node_secs + 2 * dent_secs + imeta_secs +
582 reserved_sections(sbi) + needed);
39a53e0c
JK		 583}
584
4354994f
DR		 585static inline int f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
586{
587 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
588 return 0;
589 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
590 return 0;
591 return -ENOSPC;
592}
593
81eb8d6e
JK		 594static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
595{
6c311ec6 596 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
81eb8d6e
JK		 597}
598
39a53e0c
JK		 599static inline int utilization(struct f2fs_sb_info *sbi)
600{
6c311ec6
CF		 601 return div_u64((u64)valid_user_blocks(sbi) * 100,
602 sbi->user_block_count);
39a53e0c
JK		 603}
604
605/*
606 * Sometimes f2fs may be better to drop out-of-place update policy.
216fbd64
JK		 607 * And, users can control the policy through sysfs entries.
608 * There are five policies with triggering conditions as follows.
609 * F2FS_IPU_FORCE - all the time,
610 * F2FS_IPU_SSR - if SSR mode is activated,
611 * F2FS_IPU_UTIL - if FS utilization is over threashold,
612 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
613 * threashold,
c1ce1b02
JK		 614 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
615 * storages. IPU will be triggered only if the # of dirty
616 * pages over min_fsync_blocks.
216fbd64 617 * F2FS_IPUT_DISABLE - disable IPU. (=default option)
39a53e0c 618 */
216fbd64 619#define DEF_MIN_IPU_UTIL 70
c1ce1b02 620#define DEF_MIN_FSYNC_BLOCKS 8
ef095d19 621#define DEF_MIN_HOT_BLOCKS 16
216fbd64 622
84b89e5d
JK		 623#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
624
216fbd64
JK		 625enum {
626 F2FS_IPU_FORCE,
627 F2FS_IPU_SSR,
628 F2FS_IPU_UTIL,
629 F2FS_IPU_SSR_UTIL,
c1ce1b02 630 F2FS_IPU_FSYNC,
04485987 631 F2FS_IPU_ASYNC,
216fbd64
JK		 632};
633
39a53e0c
JK		 634static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
635 int type)
636{
637 struct curseg_info *curseg = CURSEG_I(sbi, type);
638 return curseg->segno;
639}
640
641static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
642 int type)
643{
644 struct curseg_info *curseg = CURSEG_I(sbi, type);
645 return curseg->alloc_type;
646}
647
648static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
649{
650 struct curseg_info *curseg = CURSEG_I(sbi, type);
651 return curseg->next_blkoff;
652}
653
654static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
655{
7a04f64d 656 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
39a53e0c
JK		 657}
658
0833721e 659static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
39a53e0c 660{
0833721e
YH		 661 struct f2fs_sb_info *sbi = fio->sbi;
662
c9b60788 663 if (__is_meta_io(fio))
e1da7872 664 verify_blkaddr(sbi, blk_addr, META_GENERIC);
0833721e 665 else
e1da7872 666 verify_blkaddr(sbi, blk_addr, DATA_GENERIC);
39a53e0c
JK		 667}
668
669/*
e1c42045 670 * Summary block is always treated as an invalid block
39a53e0c 671 */
c39a1b34 672static inline int check_block_count(struct f2fs_sb_info *sbi,
39a53e0c
JK		 673 int segno, struct f2fs_sit_entry *raw_sit)
674{
4c278394 675#ifdef CONFIG_F2FS_CHECK_FS
44c60bf2 676 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
39a53e0c 677 int valid_blocks = 0;
44c60bf2 678 int cur_pos = 0, next_pos;
39a53e0c 679
39a53e0c 680 /* check bitmap with valid block count */
44c60bf2
CY		 681 do {
682 if (is_valid) {
683 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
684 sbi->blocks_per_seg,
685 cur_pos);
686 valid_blocks += next_pos - cur_pos;
687 } else
688 next_pos = find_next_bit_le(&raw_sit->valid_map,
689 sbi->blocks_per_seg,
690 cur_pos);
691 cur_pos = next_pos;
692 is_valid = !is_valid;
693 } while (cur_pos < sbi->blocks_per_seg);
c39a1b34
JK		 694
695 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
696 f2fs_msg(sbi->sb, KERN_ERR,
697 "Mismatch valid blocks %d vs. %d",
698 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
699 set_sbi_flag(sbi, SBI_NEED_FSCK);
700 return -EINVAL;
701 }
5d56b671 702#endif
4c278394 703 /* check segment usage, and check boundary of a given segment number */
c39a1b34
JK		 704 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
705 || segno > TOTAL_SEGS(sbi) - 1)) {
706 f2fs_msg(sbi->sb, KERN_ERR,
707 "Wrong valid blocks %d or segno %u",
708 GET_SIT_VBLOCKS(raw_sit), segno);
709 set_sbi_flag(sbi, SBI_NEED_FSCK);
710 return -EINVAL;
711 }
712 return 0;
7a04f64d 713}
39a53e0c
JK		 714
715static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
716 unsigned int start)
717{
718 struct sit_info *sit_i = SIT_I(sbi);
d3a14afd 719 unsigned int offset = SIT_BLOCK_OFFSET(start);
39a53e0c
JK		 720 block_t blk_addr = sit_i->sit_base_addr + offset;
721
722 check_seg_range(sbi, start);
723
ae27d62e
CY		 724#ifdef CONFIG_F2FS_CHECK_FS
725 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
726 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
727 f2fs_bug_on(sbi, 1);
728#endif
729
39a53e0c
JK		 730 /* calculate sit block address */
731 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
732 blk_addr += sit_i->sit_blocks;
733
734 return blk_addr;
735}
736
737static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
738 pgoff_t block_addr)
739{
740 struct sit_info *sit_i = SIT_I(sbi);
741 block_addr -= sit_i->sit_base_addr;
742 if (block_addr < sit_i->sit_blocks)
743 block_addr += sit_i->sit_blocks;
744 else
745 block_addr -= sit_i->sit_blocks;
746
747 return block_addr + sit_i->sit_base_addr;
748}
749
750static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
751{
d3a14afd 752 unsigned int block_off = SIT_BLOCK_OFFSET(start);
39a53e0c 753
c6ac4c0e 754 f2fs_change_bit(block_off, sit_i->sit_bitmap);
ae27d62e
CY		 755#ifdef CONFIG_F2FS_CHECK_FS
756 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
757#endif
39a53e0c
JK		 758}
759
a1f72ac2
CY		 760static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
761 bool base_time)
39a53e0c
JK		 762{
763 struct sit_info *sit_i = SIT_I(sbi);
a1f72ac2 764 time64_t diff, now = ktime_get_real_seconds();
48fbfe50 765
a1f72ac2
CY		 766 if (now >= sit_i->mounted_time)
767 return sit_i->elapsed_time + now - sit_i->mounted_time;
768
769 /* system time is set to the past */
770 if (!base_time) {
771 diff = sit_i->mounted_time - now;
772 if (sit_i->elapsed_time >= diff)
773 return sit_i->elapsed_time - diff;
774 return 0;
775 }
776 return sit_i->elapsed_time;
39a53e0c
JK		 777}
778
779static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
780 unsigned int ofs_in_node, unsigned char version)
781{
782 sum->nid = cpu_to_le32(nid);
783 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
784 sum->version = version;
785}
786
787static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
788{
789 return __start_cp_addr(sbi) +
790 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
791}
792
793static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
794{
795 return __start_cp_addr(sbi) +
796 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
797 - (base + 1) + type;
798}
5ec4e49f
JK		 799
800static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
801{
802 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
803 return true;
804 return false;
805}
ac5d156c 806
87d6f890
JK		 807/*
808 * It is very important to gather dirty pages and write at once, so that we can
809 * submit a big bio without interfering other data writes.
810 * By default, 512 pages for directory data,
727ebb09
KM		 811 * 512 pages (2MB) * 8 for nodes, and
812 * 256 pages * 8 for meta are set.
87d6f890
JK		 813 */
814static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
815{
a88a341a 816 if (sbi->sb->s_bdi->wb.dirty_exceeded)
510184c8
JK		 817 return 0;
818
a1257023
JK		 819 if (type == DATA)
820 return sbi->blocks_per_seg;
821 else if (type == NODE)
2c237eba 822 return 8 * sbi->blocks_per_seg;
87d6f890 823 else if (type == META)
664ba972 824 return 8 * BIO_MAX_PAGES;
87d6f890
JK		 825 else
826 return 0;
827}
50c8cdb3
JK		 828
829/*
830 * When writing pages, it'd better align nr_to_write for segment size.
831 */
832static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
833 struct writeback_control *wbc)
834{
835 long nr_to_write, desired;
836
837 if (wbc->sync_mode != WB_SYNC_NONE)
838 return 0;
839
840 nr_to_write = wbc->nr_to_write;
664ba972 841 desired = BIO_MAX_PAGES;
28ea6162 842 if (type == NODE)
664ba972 843 desired <<= 1;
50c8cdb3
JK		 844
845 wbc->nr_to_write = desired;
846 return desired - nr_to_write;
847}
01983c71
JK		 848
849static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
850{
851 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
852 bool wakeup = false;
853 int i;
854
855 if (force)
856 goto wake_up;
857
858 mutex_lock(&dcc->cmd_lock);
78997b56
CY		 859 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
860 if (i + 1 < dcc->discard_granularity)
861 break;
01983c71
JK		 862 if (!list_empty(&dcc->pend_list[i])) {
863 wakeup = true;
864 break;
865 }
866 }
867 mutex_unlock(&dcc->cmd_lock);
868 if (!wakeup)
869 return;
870wake_up:
871 dcc->discard_wake = 1;
872 wake_up_interruptible_all(&dcc->discard_wait_queue);
873}
Linux 6.x block layer and io_uring tree(s)