projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Merge tag 'pm-6.16-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
[linux-2.6-block.git] / fs / f2fs / segment.h
... / ...
CommitLineData
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_START_SEG_FROM_SEC(sbi, segno) \
106 (rounddown(segno, SEGS_PER_SEC(sbi)))
107#define GET_SEC_FROM_SEG(sbi, segno) \
108 (((segno) == -1) ? -1 : (segno) / SEGS_PER_SEC(sbi))
109#define GET_SEG_FROM_SEC(sbi, secno) \
110 ((secno) * SEGS_PER_SEC(sbi))
111#define GET_ZONE_FROM_SEC(sbi, secno) \
112 (((secno) == -1) ? -1 : (secno) / (sbi)->secs_per_zone)
113#define GET_ZONE_FROM_SEG(sbi, segno) \
114 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
115
116#define GET_SUM_BLOCK(sbi, segno) \
117 ((sbi)->sm_info->ssa_blkaddr + (segno))
118
119#define GET_SUM_TYPE(footer) ((footer)->entry_type)
120#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
121
122#define SIT_ENTRY_OFFSET(sit_i, segno) \
123 ((segno) % (sit_i)->sents_per_block)
124#define SIT_BLOCK_OFFSET(segno) \
125 ((segno) / SIT_ENTRY_PER_BLOCK)
126#define START_SEGNO(segno) \
127 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
128#define SIT_BLK_CNT(sbi) \
129 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
130#define f2fs_bitmap_size(nr) \
131 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
132
133#define SECTOR_FROM_BLOCK(blk_addr) \
134 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
135#define SECTOR_TO_BLOCK(sectors) \
136 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
137
138/*
139 * In the victim_sel_policy->alloc_mode, there are three block allocation modes.
140 * LFS writes data sequentially with cleaning operations.
141 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
142 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
143 * fragmented segment which has similar aging degree.
144 */
145enum {
146 LFS = 0,
147 SSR,
148 AT_SSR,
149};
150
151/*
152 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes.
153 * GC_CB is based on cost-benefit algorithm.
154 * GC_GREEDY is based on greedy algorithm.
155 * GC_AT is based on age-threshold algorithm.
156 */
157enum {
158 GC_CB = 0,
159 GC_GREEDY,
160 GC_AT,
161 ALLOC_NEXT,
162 FLUSH_DEVICE,
163 MAX_GC_POLICY,
164};
165
166/*
167 * BG_GC means the background cleaning job.
168 * FG_GC means the on-demand cleaning job.
169 */
170enum {
171 BG_GC = 0,
172 FG_GC,
173};
174
175/* for a function parameter to select a victim segment */
176struct victim_sel_policy {
177 int alloc_mode; /* LFS or SSR */
178 int gc_mode; /* GC_CB or GC_GREEDY */
179 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
180 unsigned int max_search; /*
181 * maximum # of segments/sections
182 * to search
183 */
184 unsigned int offset; /* last scanned bitmap offset */
185 unsigned int ofs_unit; /* bitmap search unit */
186 unsigned int min_cost; /* minimum cost */
187 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
188 unsigned int min_segno; /* segment # having min. cost */
189 unsigned long long age; /* mtime of GCed section*/
190 unsigned long long age_threshold;/* age threshold */
191 bool one_time_gc; /* one time GC */
192};
193
194struct seg_entry {
195 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
196 unsigned int valid_blocks:10; /* # of valid blocks */
197 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
198 unsigned int padding:6; /* padding */
199 unsigned char *cur_valid_map; /* validity bitmap of blocks */
200#ifdef CONFIG_F2FS_CHECK_FS
201 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
202#endif
203 /*
204 * # of valid blocks and the validity bitmap stored in the last
205 * checkpoint pack. This information is used by the SSR mode.
206 */
207 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
208 unsigned char *discard_map;
209 unsigned long long mtime; /* modification time of the segment */
210};
211
212struct sec_entry {
213 unsigned int valid_blocks; /* # of valid blocks in a section */
214 unsigned int ckpt_valid_blocks; /* # of valid blocks last cp in a section */
215};
216
217#define MAX_SKIP_GC_COUNT 16
218
219struct revoke_entry {
220 struct list_head list;
221 block_t old_addr; /* for revoking when fail to commit */
222 pgoff_t index;
223};
224
225struct sit_info {
226 block_t sit_base_addr; /* start block address of SIT area */
227 block_t sit_blocks; /* # of blocks used by SIT area */
228 block_t written_valid_blocks; /* # of valid blocks in main area */
229 char *bitmap; /* all bitmaps pointer */
230 char *sit_bitmap; /* SIT bitmap pointer */
231#ifdef CONFIG_F2FS_CHECK_FS
232 char *sit_bitmap_mir; /* SIT bitmap mirror */
233
234 /* bitmap of segments to be ignored by GC in case of errors */
235 unsigned long *invalid_segmap;
236#endif
237 unsigned int bitmap_size; /* SIT bitmap size */
238
239 unsigned long *tmp_map; /* bitmap for temporal use */
240 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
241 unsigned int dirty_sentries; /* # of dirty sentries */
242 unsigned int sents_per_block; /* # of SIT entries per block */
243 struct rw_semaphore sentry_lock; /* to protect SIT cache */
244 struct seg_entry *sentries; /* SIT segment-level cache */
245 struct sec_entry *sec_entries; /* SIT section-level cache */
246
247 /* for cost-benefit algorithm in cleaning procedure */
248 unsigned long long elapsed_time; /* elapsed time after mount */
249 unsigned long long mounted_time; /* mount time */
250 unsigned long long min_mtime; /* min. modification time */
251 unsigned long long max_mtime; /* max. modification time */
252 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
253 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
254
255 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
256};
257
258struct free_segmap_info {
259 unsigned int start_segno; /* start segment number logically */
260 unsigned int free_segments; /* # of free segments */
261 unsigned int free_sections; /* # of free sections */
262 spinlock_t segmap_lock; /* free segmap lock */
263 unsigned long *free_segmap; /* free segment bitmap */
264 unsigned long *free_secmap; /* free section bitmap */
265};
266
267/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
268enum dirty_type {
269 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
270 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
271 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
272 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
273 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
274 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
275 DIRTY, /* to count # of dirty segments */
276 PRE, /* to count # of entirely obsolete segments */
277 NR_DIRTY_TYPE
278};
279
280struct dirty_seglist_info {
281 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
282 unsigned long *dirty_secmap;
283 struct mutex seglist_lock; /* lock for segment bitmaps */
284 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
285 unsigned long *victim_secmap; /* background GC victims */
286 unsigned long *pinned_secmap; /* pinned victims from foreground GC */
287 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */
288 bool enable_pin_section; /* enable pinning section */
289};
290
291/* for active log information */
292struct curseg_info {
293 struct mutex curseg_mutex; /* lock for consistency */
294 struct f2fs_summary_block *sum_blk; /* cached summary block */
295 struct rw_semaphore journal_rwsem; /* protect journal area */
296 struct f2fs_journal *journal; /* cached journal info */
297 unsigned char alloc_type; /* current allocation type */
298 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
299 unsigned int segno; /* current segment number */
300 unsigned short next_blkoff; /* next block offset to write */
301 unsigned int zone; /* current zone number */
302 unsigned int next_segno; /* preallocated segment */
303 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */
304 bool inited; /* indicate inmem log is inited */
305};
306
307struct sit_entry_set {
308 struct list_head set_list; /* link with all sit sets */
309 unsigned int start_segno; /* start segno of sits in set */
310 unsigned int entry_cnt; /* the # of sit entries in set */
311};
312
313/*
314 * inline functions
315 */
316static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
317{
318 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
319}
320
321static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
322 unsigned int segno)
323{
324 struct sit_info *sit_i = SIT_I(sbi);
325 return &sit_i->sentries[segno];
326}
327
328static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
329 unsigned int segno)
330{
331 struct sit_info *sit_i = SIT_I(sbi);
332 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
333}
334
335static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
336 unsigned int segno, bool use_section)
337{
338 /*
339 * In order to get # of valid blocks in a section instantly from many
340 * segments, f2fs manages two counting structures separately.
341 */
342 if (use_section && __is_large_section(sbi))
343 return get_sec_entry(sbi, segno)->valid_blocks;
344 else
345 return get_seg_entry(sbi, segno)->valid_blocks;
346}
347
348static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
349 unsigned int segno, bool use_section)
350{
351 if (use_section && __is_large_section(sbi))
352 return get_sec_entry(sbi, segno)->ckpt_valid_blocks;
353 else
354 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
355}
356
357static inline void set_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
358 unsigned int segno)
359{
360 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
361 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
362 unsigned int blocks = 0;
363 int i;
364
365 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) {
366 struct seg_entry *se = get_seg_entry(sbi, start_segno);
367
368 blocks += se->ckpt_valid_blocks;
369 }
370 get_sec_entry(sbi, segno)->ckpt_valid_blocks = blocks;
371}
372
373#ifdef CONFIG_F2FS_CHECK_FS
374static inline void sanity_check_valid_blocks(struct f2fs_sb_info *sbi,
375 unsigned int segno)
376{
377 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
378 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
379 unsigned int blocks = 0;
380 int i;
381
382 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) {
383 struct seg_entry *se = get_seg_entry(sbi, start_segno);
384
385 blocks += se->ckpt_valid_blocks;
386 }
387
388 if (blocks != get_sec_entry(sbi, segno)->ckpt_valid_blocks) {
389 f2fs_err(sbi,
390 "Inconsistent ckpt valid blocks: "
391 "seg entry(%d) vs sec entry(%d) at secno %d",
392 blocks, get_sec_entry(sbi, segno)->ckpt_valid_blocks, secno);
393 f2fs_bug_on(sbi, 1);
394 }
395}
396#else
397static inline void sanity_check_valid_blocks(struct f2fs_sb_info *sbi,
398 unsigned int segno)
399{
400}
401#endif
402static inline void seg_info_from_raw_sit(struct seg_entry *se,
403 struct f2fs_sit_entry *rs)
404{
405 se->valid_blocks = GET_SIT_VBLOCKS(rs);
406 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
407 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
408 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
409#ifdef CONFIG_F2FS_CHECK_FS
410 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
411#endif
412 se->type = GET_SIT_TYPE(rs);
413 se->mtime = le64_to_cpu(rs->mtime);
414}
415
416static inline void __seg_info_to_raw_sit(struct seg_entry *se,
417 struct f2fs_sit_entry *rs)
418{
419 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
420 se->valid_blocks;
421 rs->vblocks = cpu_to_le16(raw_vblocks);
422 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
423 rs->mtime = cpu_to_le64(se->mtime);
424}
425
426static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi,
427 struct folio *folio, unsigned int start)
428{
429 struct f2fs_sit_block *raw_sit;
430 struct seg_entry *se;
431 struct f2fs_sit_entry *rs;
432 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
433 (unsigned long)MAIN_SEGS(sbi));
434 int i;
435
436 raw_sit = folio_address(folio);
437 memset(raw_sit, 0, PAGE_SIZE);
438 for (i = 0; i < end - start; i++) {
439 rs = &raw_sit->entries[i];
440 se = get_seg_entry(sbi, start + i);
441 __seg_info_to_raw_sit(se, rs);
442 }
443}
444
445static inline void seg_info_to_raw_sit(struct seg_entry *se,
446 struct f2fs_sit_entry *rs)
447{
448 __seg_info_to_raw_sit(se, rs);
449
450 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
451 se->ckpt_valid_blocks = se->valid_blocks;
452}
453
454static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
455 unsigned int max, unsigned int segno)
456{
457 unsigned int ret;
458 spin_lock(&free_i->segmap_lock);
459 ret = find_next_bit(free_i->free_segmap, max, segno);
460 spin_unlock(&free_i->segmap_lock);
461 return ret;
462}
463
464static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
465{
466 struct free_segmap_info *free_i = FREE_I(sbi);
467 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
468 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
469 unsigned int next;
470
471 spin_lock(&free_i->segmap_lock);
472 clear_bit(segno, free_i->free_segmap);
473 free_i->free_segments++;
474
475 next = find_next_bit(free_i->free_segmap,
476 start_segno + SEGS_PER_SEC(sbi), start_segno);
477 if (next >= start_segno + f2fs_usable_segs_in_sec(sbi)) {
478 clear_bit(secno, free_i->free_secmap);
479 free_i->free_sections++;
480 }
481 spin_unlock(&free_i->segmap_lock);
482}
483
484static inline void __set_inuse(struct f2fs_sb_info *sbi,
485 unsigned int segno)
486{
487 struct free_segmap_info *free_i = FREE_I(sbi);
488 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
489
490 set_bit(segno, free_i->free_segmap);
491 free_i->free_segments--;
492 if (!test_and_set_bit(secno, free_i->free_secmap))
493 free_i->free_sections--;
494}
495
496static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
497 unsigned int segno, bool inmem)
498{
499 struct free_segmap_info *free_i = FREE_I(sbi);
500 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
501 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
502 unsigned int next;
503 bool ret;
504
505 spin_lock(&free_i->segmap_lock);
506 ret = test_and_clear_bit(segno, free_i->free_segmap);
507 if (!ret)
508 goto unlock_out;
509
510 free_i->free_segments++;
511
512 if (!inmem && IS_CURSEC(sbi, secno))
513 goto unlock_out;
514
515 /* check large section */
516 next = find_next_bit(free_i->free_segmap,
517 start_segno + SEGS_PER_SEC(sbi), start_segno);
518 if (next < start_segno + f2fs_usable_segs_in_sec(sbi))
519 goto unlock_out;
520
521 ret = test_and_clear_bit(secno, free_i->free_secmap);
522 if (!ret)
523 goto unlock_out;
524
525 free_i->free_sections++;
526
527 if (GET_SEC_FROM_SEG(sbi, sbi->next_victim_seg[BG_GC]) == secno)
528 sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
529 if (GET_SEC_FROM_SEG(sbi, sbi->next_victim_seg[FG_GC]) == secno)
530 sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
531
532unlock_out:
533 spin_unlock(&free_i->segmap_lock);
534}
535
536static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
537 unsigned int segno)
538{
539 struct free_segmap_info *free_i = FREE_I(sbi);
540 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
541
542 spin_lock(&free_i->segmap_lock);
543 if (!test_and_set_bit(segno, free_i->free_segmap)) {
544 free_i->free_segments--;
545 if (!test_and_set_bit(secno, free_i->free_secmap))
546 free_i->free_sections--;
547 }
548 spin_unlock(&free_i->segmap_lock);
549}
550
551static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
552 void *dst_addr)
553{
554 struct sit_info *sit_i = SIT_I(sbi);
555
556#ifdef CONFIG_F2FS_CHECK_FS
557 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
558 sit_i->bitmap_size))
559 f2fs_bug_on(sbi, 1);
560#endif
561 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
562}
563
564static inline block_t written_block_count(struct f2fs_sb_info *sbi)
565{
566 return SIT_I(sbi)->written_valid_blocks;
567}
568
569static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
570{
571 return FREE_I(sbi)->free_segments;
572}
573
574static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
575{
576 return SM_I(sbi)->reserved_segments;
577}
578
579static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
580{
581 return FREE_I(sbi)->free_sections;
582}
583
584static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
585{
586 return DIRTY_I(sbi)->nr_dirty[PRE];
587}
588
589static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
590{
591 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
592 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
593 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
594 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
595 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
596 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
597}
598
599static inline int overprovision_segments(struct f2fs_sb_info *sbi)
600{
601 return SM_I(sbi)->ovp_segments;
602}
603
604static inline int reserved_sections(struct f2fs_sb_info *sbi)
605{
606 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
607}
608
609static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
610 unsigned int node_blocks, unsigned int data_blocks,
611 unsigned int dent_blocks)
612{
613 unsigned int segno, left_blocks, blocks;
614 int i;
615
616 /* check current data/node sections in the worst case. */
617 for (i = CURSEG_HOT_DATA; i < NR_PERSISTENT_LOG; i++) {
618 segno = CURSEG_I(sbi, i)->segno;
619
620 if (unlikely(segno == NULL_SEGNO))
621 return false;
622
623 if (f2fs_lfs_mode(sbi) && __is_large_section(sbi)) {
624 left_blocks = CAP_BLKS_PER_SEC(sbi) -
625 SEGS_TO_BLKS(sbi, (segno - GET_START_SEG_FROM_SEC(sbi, segno))) -
626 CURSEG_I(sbi, i)->next_blkoff;
627 } else {
628 left_blocks = CAP_BLKS_PER_SEC(sbi) -
629 get_ckpt_valid_blocks(sbi, segno, true);
630 }
631
632 blocks = i <= CURSEG_COLD_DATA ? data_blocks : node_blocks;
633 if (blocks > left_blocks)
634 return false;
635 }
636
637 /* check current data section for dentry blocks. */
638 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
639
640 if (unlikely(segno == NULL_SEGNO))
641 return false;
642
643 if (f2fs_lfs_mode(sbi) && __is_large_section(sbi)) {
644 left_blocks = CAP_BLKS_PER_SEC(sbi) -
645 SEGS_TO_BLKS(sbi, (segno - GET_START_SEG_FROM_SEC(sbi, segno))) -
646 CURSEG_I(sbi, CURSEG_HOT_DATA)->next_blkoff;
647 } else {
648 left_blocks = CAP_BLKS_PER_SEC(sbi) -
649 get_ckpt_valid_blocks(sbi, segno, true);
650 }
651
652 if (dent_blocks > left_blocks)
653 return false;
654 return true;
655}
656
657/*
658 * calculate needed sections for dirty node/dentry and call
659 * has_curseg_enough_space, please note that, it needs to account
660 * dirty data as well in lfs mode when checkpoint is disabled.
661 */
662static inline void __get_secs_required(struct f2fs_sb_info *sbi,
663 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p)
664{
665 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
666 get_pages(sbi, F2FS_DIRTY_DENTS) +
667 get_pages(sbi, F2FS_DIRTY_IMETA);
668 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
669 unsigned int total_data_blocks = 0;
670 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
671 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
672 unsigned int data_secs = 0;
673 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
674 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
675 unsigned int data_blocks = 0;
676
677 if (f2fs_lfs_mode(sbi) &&
678 unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
679 total_data_blocks = get_pages(sbi, F2FS_DIRTY_DATA);
680 data_secs = total_data_blocks / CAP_BLKS_PER_SEC(sbi);
681 data_blocks = total_data_blocks % CAP_BLKS_PER_SEC(sbi);
682 }
683
684 if (lower_p)
685 *lower_p = node_secs + dent_secs + data_secs;
686 if (upper_p)
687 *upper_p = node_secs + dent_secs +
688 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0) +
689 (data_blocks ? 1 : 0);
690 if (curseg_p)
691 *curseg_p = has_curseg_enough_space(sbi,
692 node_blocks, data_blocks, dent_blocks);
693}
694
695static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
696 int freed, int needed)
697{
698 unsigned int free_secs, lower_secs, upper_secs;
699 bool curseg_space;
700
701 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
702 return false;
703
704 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space);
705
706 free_secs = free_sections(sbi) + freed;
707 lower_secs += needed + reserved_sections(sbi);
708 upper_secs += needed + reserved_sections(sbi);
709
710 if (free_secs > upper_secs)
711 return false;
712 if (free_secs <= lower_secs)
713 return true;
714 return !curseg_space;
715}
716
717static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi,
718 int freed, int needed)
719{
720 return !has_not_enough_free_secs(sbi, freed, needed);
721}
722
723static inline bool has_enough_free_blks(struct f2fs_sb_info *sbi)
724{
725 unsigned int total_free_blocks = 0;
726 unsigned int avail_user_block_count;
727
728 spin_lock(&sbi->stat_lock);
729
730 avail_user_block_count = get_available_block_count(sbi, NULL, true);
731 total_free_blocks = avail_user_block_count - (unsigned int)valid_user_blocks(sbi);
732
733 spin_unlock(&sbi->stat_lock);
734
735 return total_free_blocks > 0;
736}
737
738static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
739{
740 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
741 return true;
742 if (likely(has_enough_free_secs(sbi, 0, 0)))
743 return true;
744 if (!f2fs_lfs_mode(sbi) &&
745 likely(has_enough_free_blks(sbi)))
746 return true;
747 return false;
748}
749
750static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
751{
752 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
753}
754
755static inline int utilization(struct f2fs_sb_info *sbi)
756{
757 return div_u64((u64)valid_user_blocks(sbi) * 100,
758 sbi->user_block_count);
759}
760
761/*
762 * Sometimes f2fs may be better to drop out-of-place update policy.
763 * And, users can control the policy through sysfs entries.
764 * There are five policies with triggering conditions as follows.
765 * F2FS_IPU_FORCE - all the time,
766 * F2FS_IPU_SSR - if SSR mode is activated,
767 * F2FS_IPU_UTIL - if FS utilization is over threashold,
768 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
769 * threashold,
770 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
771 * storages. IPU will be triggered only if the # of dirty
772 * pages over min_fsync_blocks. (=default option)
773 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
774 * F2FS_IPU_NOCACHE - disable IPU bio cache.
775 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
776 * FI_OPU_WRITE flag.
777 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
778 */
779#define DEF_MIN_IPU_UTIL 70
780#define DEF_MIN_FSYNC_BLOCKS 8
781#define DEF_MIN_HOT_BLOCKS 16
782
783#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
784
785#define F2FS_IPU_DISABLE 0
786
787/* Modification on enum should be synchronized with ipu_mode_names array */
788enum {
789 F2FS_IPU_FORCE,
790 F2FS_IPU_SSR,
791 F2FS_IPU_UTIL,
792 F2FS_IPU_SSR_UTIL,
793 F2FS_IPU_FSYNC,
794 F2FS_IPU_ASYNC,
795 F2FS_IPU_NOCACHE,
796 F2FS_IPU_HONOR_OPU_WRITE,
797 F2FS_IPU_MAX,
798};
799
800static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi)
801{
802 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE;
803}
804
805#define F2FS_IPU_POLICY(name) \
806static inline bool IS_##name(struct f2fs_sb_info *sbi) \
807{ \
808 return SM_I(sbi)->ipu_policy & BIT(name); \
809}
810
811F2FS_IPU_POLICY(F2FS_IPU_FORCE);
812F2FS_IPU_POLICY(F2FS_IPU_SSR);
813F2FS_IPU_POLICY(F2FS_IPU_UTIL);
814F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL);
815F2FS_IPU_POLICY(F2FS_IPU_FSYNC);
816F2FS_IPU_POLICY(F2FS_IPU_ASYNC);
817F2FS_IPU_POLICY(F2FS_IPU_NOCACHE);
818F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE);
819
820static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
821 int type)
822{
823 struct curseg_info *curseg = CURSEG_I(sbi, type);
824 return curseg->segno;
825}
826
827static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
828 int type)
829{
830 struct curseg_info *curseg = CURSEG_I(sbi, type);
831 return curseg->alloc_type;
832}
833
834static inline bool valid_main_segno(struct f2fs_sb_info *sbi,
835 unsigned int segno)
836{
837 return segno <= (MAIN_SEGS(sbi) - 1);
838}
839
840static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
841{
842 struct f2fs_sb_info *sbi = fio->sbi;
843
844 if (__is_valid_data_blkaddr(fio->old_blkaddr))
845 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
846 META_GENERIC : DATA_GENERIC);
847 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
848 META_GENERIC : DATA_GENERIC_ENHANCE);
849}
850
851/*
852 * Summary block is always treated as an invalid block
853 */
854static inline int check_block_count(struct f2fs_sb_info *sbi,
855 int segno, struct f2fs_sit_entry *raw_sit)
856{
857 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
858 int valid_blocks = 0;
859 int cur_pos = 0, next_pos;
860 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
861
862 /* check bitmap with valid block count */
863 do {
864 if (is_valid) {
865 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
866 usable_blks_per_seg,
867 cur_pos);
868 valid_blocks += next_pos - cur_pos;
869 } else
870 next_pos = find_next_bit_le(&raw_sit->valid_map,
871 usable_blks_per_seg,
872 cur_pos);
873 cur_pos = next_pos;
874 is_valid = !is_valid;
875 } while (cur_pos < usable_blks_per_seg);
876
877 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
878 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
879 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
880 set_sbi_flag(sbi, SBI_NEED_FSCK);
881 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
882 return -EFSCORRUPTED;
883 }
884
885 if (usable_blks_per_seg < BLKS_PER_SEG(sbi))
886 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
887 BLKS_PER_SEG(sbi),
888 usable_blks_per_seg) != BLKS_PER_SEG(sbi));
889
890 /* check segment usage, and check boundary of a given segment number */
891 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
892 || !valid_main_segno(sbi, segno))) {
893 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
894 GET_SIT_VBLOCKS(raw_sit), segno);
895 set_sbi_flag(sbi, SBI_NEED_FSCK);
896 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
897 return -EFSCORRUPTED;
898 }
899 return 0;
900}
901
902static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
903 unsigned int start)
904{
905 struct sit_info *sit_i = SIT_I(sbi);
906 unsigned int offset = SIT_BLOCK_OFFSET(start);
907 block_t blk_addr = sit_i->sit_base_addr + offset;
908
909 f2fs_bug_on(sbi, !valid_main_segno(sbi, start));
910
911#ifdef CONFIG_F2FS_CHECK_FS
912 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
913 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
914 f2fs_bug_on(sbi, 1);
915#endif
916
917 /* calculate sit block address */
918 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
919 blk_addr += sit_i->sit_blocks;
920
921 return blk_addr;
922}
923
924static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
925 pgoff_t block_addr)
926{
927 struct sit_info *sit_i = SIT_I(sbi);
928 block_addr -= sit_i->sit_base_addr;
929 if (block_addr < sit_i->sit_blocks)
930 block_addr += sit_i->sit_blocks;
931 else
932 block_addr -= sit_i->sit_blocks;
933
934 return block_addr + sit_i->sit_base_addr;
935}
936
937static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
938{
939 unsigned int block_off = SIT_BLOCK_OFFSET(start);
940
941 f2fs_change_bit(block_off, sit_i->sit_bitmap);
942#ifdef CONFIG_F2FS_CHECK_FS
943 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
944#endif
945}
946
947static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
948 bool base_time)
949{
950 struct sit_info *sit_i = SIT_I(sbi);
951 time64_t diff, now = ktime_get_boottime_seconds();
952
953 if (now >= sit_i->mounted_time)
954 return sit_i->elapsed_time + now - sit_i->mounted_time;
955
956 /* system time is set to the past */
957 if (!base_time) {
958 diff = sit_i->mounted_time - now;
959 if (sit_i->elapsed_time >= diff)
960 return sit_i->elapsed_time - diff;
961 return 0;
962 }
963 return sit_i->elapsed_time;
964}
965
966static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
967 unsigned int ofs_in_node, unsigned char version)
968{
969 sum->nid = cpu_to_le32(nid);
970 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
971 sum->version = version;
972}
973
974static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
975{
976 return __start_cp_addr(sbi) +
977 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
978}
979
980static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
981{
982 return __start_cp_addr(sbi) +
983 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
984 - (base + 1) + type;
985}
986
987static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
988{
989 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
990 return true;
991 return false;
992}
993
994/*
995 * It is very important to gather dirty pages and write at once, so that we can
996 * submit a big bio without interfering other data writes.
997 * By default, 512 pages for directory data,
998 * 512 pages (2MB) * 8 for nodes, and
999 * 256 pages * 8 for meta are set.
1000 */
1001static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
1002{
1003 if (sbi->sb->s_bdi->wb.dirty_exceeded)
1004 return 0;
1005
1006 if (type == DATA)
1007 return BLKS_PER_SEG(sbi);
1008 else if (type == NODE)
1009 return SEGS_TO_BLKS(sbi, 8);
1010 else if (type == META)
1011 return 8 * BIO_MAX_VECS;
1012 else
1013 return 0;
1014}
1015
1016/*
1017 * When writing pages, it'd better align nr_to_write for segment size.
1018 */
1019static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
1020 struct writeback_control *wbc)
1021{
1022 long nr_to_write, desired;
1023
1024 if (wbc->sync_mode != WB_SYNC_NONE)
1025 return 0;
1026
1027 nr_to_write = wbc->nr_to_write;
1028 desired = BIO_MAX_VECS;
1029 if (type == NODE)
1030 desired <<= 1;
1031
1032 wbc->nr_to_write = desired;
1033 return desired - nr_to_write;
1034}
1035
1036static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
1037{
1038 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1039 bool wakeup = false;
1040 int i;
1041
1042 if (force)
1043 goto wake_up;
1044
1045 mutex_lock(&dcc->cmd_lock);
1046 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1047 if (i + 1 < dcc->discard_granularity)
1048 break;
1049 if (!list_empty(&dcc->pend_list[i])) {
1050 wakeup = true;
1051 break;
1052 }
1053 }
1054 mutex_unlock(&dcc->cmd_lock);
1055 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
1056 return;
1057wake_up:
1058 dcc->discard_wake = true;
1059 wake_up_interruptible_all(&dcc->discard_wait_queue);
1060}
Linux 6.x block layer and io_uring tree(s)