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 /* L: Logical segment # in volume, R: Relative segment # in main area */
22 #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno)
23 #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno)
24
25 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA)
26 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
27 #define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA))
28
29 static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
30 unsigned short seg_type)
31 {
32 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
33 }
34
35 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
36 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
37 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
38
39 #define IS_CURSEG(sbi, seg) \
40 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
41 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
42 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
43 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
44 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
45 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \
46 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \
47 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
48
49 #define IS_CURSEC(sbi, secno) \
50 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
51 SEGS_PER_SEC(sbi)) || \
52 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
53 SEGS_PER_SEC(sbi)) || \
54 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
55 SEGS_PER_SEC(sbi)) || \
56 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
57 SEGS_PER_SEC(sbi)) || \
58 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
59 SEGS_PER_SEC(sbi)) || \
60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
61 SEGS_PER_SEC(sbi)) || \
62 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
63 SEGS_PER_SEC(sbi)) || \
64 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
65 SEGS_PER_SEC(sbi)))
66
67 #define MAIN_BLKADDR(sbi) \
68 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
69 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
70 #define SEG0_BLKADDR(sbi) \
71 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
72 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
73
74 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
75 #define MAIN_SECS(sbi) ((sbi)->total_sections)
76
77 #define TOTAL_SEGS(sbi) \
78 (SM_I(sbi) ? SM_I(sbi)->segment_count : \
79 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
80 #define TOTAL_BLKS(sbi) (SEGS_TO_BLKS(sbi, TOTAL_SEGS(sbi)))
81
82 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
83 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \
84 (sbi)->log_blocks_per_seg))
85
86 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
87 (SEGS_TO_BLKS(sbi, GET_R2L_SEGNO(FREE_I(sbi), segno))))
88
89 #define NEXT_FREE_BLKADDR(sbi, curseg) \
90 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
91
92 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
93 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
94 (BLKS_TO_SEGS(sbi, GET_SEGOFF_FROM_SEG0(sbi, blk_addr)))
95 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
96 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (BLKS_PER_SEG(sbi) - 1))
97
98 #define GET_SEGNO(sbi, blk_addr) \
99 ((!__is_valid_data_blkaddr(blk_addr)) ? \
100 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
101 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
102 #define CAP_BLKS_PER_SEC(sbi) \
103 (BLKS_PER_SEC(sbi) - (sbi)->unusable_blocks_per_sec)
104 #define CAP_SEGS_PER_SEC(sbi) \
105 (SEGS_PER_SEC(sbi) - \
106 BLKS_TO_SEGS(sbi, (sbi)->unusable_blocks_per_sec))
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 */
145 enum {
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 */
157 enum {
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 */
170 enum {
171 BG_GC = 0,
172 FG_GC,
173 };
174
175 /* for a function parameter to select a victim segment */
176 struct 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 };
192
193 struct seg_entry {
194 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
195 unsigned int valid_blocks:10; /* # of valid blocks */
196 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
197 unsigned int padding:6; /* padding */
198 unsigned char *cur_valid_map; /* validity bitmap of blocks */
199 #ifdef CONFIG_F2FS_CHECK_FS
200 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
201 #endif
202 /*
203 * # of valid blocks and the validity bitmap stored in the last
204 * checkpoint pack. This information is used by the SSR mode.
205 */
206 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
207 unsigned char *discard_map;
208 unsigned long long mtime; /* modification time of the segment */
209 };
210
211 struct sec_entry {
212 unsigned int valid_blocks; /* # of valid blocks in a section */
213 };
214
215 #define MAX_SKIP_GC_COUNT 16
216
217 struct revoke_entry {
218 struct list_head list;
219 block_t old_addr; /* for revoking when fail to commit */
220 pgoff_t index;
221 };
222
223 struct sit_info {
224 block_t sit_base_addr; /* start block address of SIT area */
225 block_t sit_blocks; /* # of blocks used by SIT area */
226 block_t written_valid_blocks; /* # of valid blocks in main area */
227 char *bitmap; /* all bitmaps pointer */
228 char *sit_bitmap; /* SIT bitmap pointer */
229 #ifdef CONFIG_F2FS_CHECK_FS
230 char *sit_bitmap_mir; /* SIT bitmap mirror */
231
232 /* bitmap of segments to be ignored by GC in case of errors */
233 unsigned long *invalid_segmap;
234 #endif
235 unsigned int bitmap_size; /* SIT bitmap size */
236
237 unsigned long *tmp_map; /* bitmap for temporal use */
238 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
239 unsigned int dirty_sentries; /* # of dirty sentries */
240 unsigned int sents_per_block; /* # of SIT entries per block */
241 struct rw_semaphore sentry_lock; /* to protect SIT cache */
242 struct seg_entry *sentries; /* SIT segment-level cache */
243 struct sec_entry *sec_entries; /* SIT section-level cache */
244
245 /* for cost-benefit algorithm in cleaning procedure */
246 unsigned long long elapsed_time; /* elapsed time after mount */
247 unsigned long long mounted_time; /* mount time */
248 unsigned long long min_mtime; /* min. modification time */
249 unsigned long long max_mtime; /* max. modification time */
250 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
251 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
252
253 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
254 };
255
256 struct free_segmap_info {
257 unsigned int start_segno; /* start segment number logically */
258 unsigned int free_segments; /* # of free segments */
259 unsigned int free_sections; /* # of free sections */
260 spinlock_t segmap_lock; /* free segmap lock */
261 unsigned long *free_segmap; /* free segment bitmap */
262 unsigned long *free_secmap; /* free section bitmap */
263 };
264
265 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
266 enum dirty_type {
267 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
268 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
269 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
270 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
271 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
272 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
273 DIRTY, /* to count # of dirty segments */
274 PRE, /* to count # of entirely obsolete segments */
275 NR_DIRTY_TYPE
276 };
277
278 struct dirty_seglist_info {
279 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
280 unsigned long *dirty_secmap;
281 struct mutex seglist_lock; /* lock for segment bitmaps */
282 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
283 unsigned long *victim_secmap; /* background GC victims */
284 unsigned long *pinned_secmap; /* pinned victims from foreground GC */
285 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */
286 bool enable_pin_section; /* enable pinning section */
287 };
288
289 /* for active log information */
290 struct 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 short seg_type; /* segment type like CURSEG_XXX_TYPE */
297 unsigned int segno; /* current segment number */
298 unsigned short next_blkoff; /* next block offset to write */
299 unsigned int zone; /* current zone number */
300 unsigned int next_segno; /* preallocated segment */
301 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */
302 bool inited; /* indicate inmem log is inited */
303 };
304
305 struct sit_entry_set {
306 struct list_head set_list; /* link with all sit sets */
307 unsigned int start_segno; /* start segno of sits in set */
308 unsigned int entry_cnt; /* the # of sit entries in set */
309 };
310
311 /*
312 * inline functions
313 */
314 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
315 {
316 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
317 }
318
319 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
320 unsigned int segno)
321 {
322 struct sit_info *sit_i = SIT_I(sbi);
323 return &sit_i->sentries[segno];
324 }
325
326 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
327 unsigned int segno)
328 {
329 struct sit_info *sit_i = SIT_I(sbi);
330 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
331 }
332
333 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
334 unsigned int segno, bool use_section)
335 {
336 /*
337 * In order to get # of valid blocks in a section instantly from many
338 * segments, f2fs manages two counting structures separately.
339 */
340 if (use_section && __is_large_section(sbi))
341 return get_sec_entry(sbi, segno)->valid_blocks;
342 else
343 return get_seg_entry(sbi, segno)->valid_blocks;
344 }
345
346 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
347 unsigned int segno, bool use_section)
348 {
349 if (use_section && __is_large_section(sbi)) {
350 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
351 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
352 unsigned int blocks = 0;
353 int i;
354
355 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) {
356 struct seg_entry *se = get_seg_entry(sbi, start_segno);
357
358 blocks += se->ckpt_valid_blocks;
359 }
360 return blocks;
361 }
362 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
363 }
364
365 static inline void seg_info_from_raw_sit(struct seg_entry *se,
366 struct f2fs_sit_entry *rs)
367 {
368 se->valid_blocks = GET_SIT_VBLOCKS(rs);
369 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
370 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
371 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
372 #ifdef CONFIG_F2FS_CHECK_FS
373 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
374 #endif
375 se->type = GET_SIT_TYPE(rs);
376 se->mtime = le64_to_cpu(rs->mtime);
377 }
378
379 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
380 struct f2fs_sit_entry *rs)
381 {
382 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
383 se->valid_blocks;
384 rs->vblocks = cpu_to_le16(raw_vblocks);
385 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
386 rs->mtime = cpu_to_le64(se->mtime);
387 }
388
389 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
390 struct page *page, unsigned int start)
391 {
392 struct f2fs_sit_block *raw_sit;
393 struct seg_entry *se;
394 struct f2fs_sit_entry *rs;
395 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
396 (unsigned long)MAIN_SEGS(sbi));
397 int i;
398
399 raw_sit = (struct f2fs_sit_block *)page_address(page);
400 memset(raw_sit, 0, PAGE_SIZE);
401 for (i = 0; i < end - start; i++) {
402 rs = &raw_sit->entries[i];
403 se = get_seg_entry(sbi, start + i);
404 __seg_info_to_raw_sit(se, rs);
405 }
406 }
407
408 static inline void seg_info_to_raw_sit(struct seg_entry *se,
409 struct f2fs_sit_entry *rs)
410 {
411 __seg_info_to_raw_sit(se, rs);
412
413 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
414 se->ckpt_valid_blocks = se->valid_blocks;
415 }
416
417 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
418 unsigned int max, unsigned int segno)
419 {
420 unsigned int ret;
421 spin_lock(&free_i->segmap_lock);
422 ret = find_next_bit(free_i->free_segmap, max, segno);
423 spin_unlock(&free_i->segmap_lock);
424 return ret;
425 }
426
427 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
428 {
429 struct free_segmap_info *free_i = FREE_I(sbi);
430 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
431 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
432 unsigned int next;
433 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
434
435 spin_lock(&free_i->segmap_lock);
436 clear_bit(segno, free_i->free_segmap);
437 free_i->free_segments++;
438
439 next = find_next_bit(free_i->free_segmap,
440 start_segno + SEGS_PER_SEC(sbi), start_segno);
441 if (next >= start_segno + usable_segs) {
442 clear_bit(secno, free_i->free_secmap);
443 free_i->free_sections++;
444 }
445 spin_unlock(&free_i->segmap_lock);
446 }
447
448 static inline void __set_inuse(struct f2fs_sb_info *sbi,
449 unsigned int segno)
450 {
451 struct free_segmap_info *free_i = FREE_I(sbi);
452 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
453
454 set_bit(segno, free_i->free_segmap);
455 free_i->free_segments--;
456 if (!test_and_set_bit(secno, free_i->free_secmap))
457 free_i->free_sections--;
458 }
459
460 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
461 unsigned int segno, bool inmem)
462 {
463 struct free_segmap_info *free_i = FREE_I(sbi);
464 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
465 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
466 unsigned int next;
467 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
468
469 spin_lock(&free_i->segmap_lock);
470 if (test_and_clear_bit(segno, free_i->free_segmap)) {
471 free_i->free_segments++;
472
473 if (!inmem && IS_CURSEC(sbi, secno))
474 goto skip_free;
475 next = find_next_bit(free_i->free_segmap,
476 start_segno + SEGS_PER_SEC(sbi), start_segno);
477 if (next >= start_segno + usable_segs) {
478 if (test_and_clear_bit(secno, free_i->free_secmap))
479 free_i->free_sections++;
480 }
481 }
482 skip_free:
483 spin_unlock(&free_i->segmap_lock);
484 }
485
486 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
487 unsigned int segno)
488 {
489 struct free_segmap_info *free_i = FREE_I(sbi);
490 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
491
492 spin_lock(&free_i->segmap_lock);
493 if (!test_and_set_bit(segno, free_i->free_segmap)) {
494 free_i->free_segments--;
495 if (!test_and_set_bit(secno, free_i->free_secmap))
496 free_i->free_sections--;
497 }
498 spin_unlock(&free_i->segmap_lock);
499 }
500
501 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
502 void *dst_addr)
503 {
504 struct sit_info *sit_i = SIT_I(sbi);
505
506 #ifdef CONFIG_F2FS_CHECK_FS
507 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
508 sit_i->bitmap_size))
509 f2fs_bug_on(sbi, 1);
510 #endif
511 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
512 }
513
514 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
515 {
516 return SIT_I(sbi)->written_valid_blocks;
517 }
518
519 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
520 {
521 return FREE_I(sbi)->free_segments;
522 }
523
524 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
525 {
526 return SM_I(sbi)->reserved_segments +
527 SM_I(sbi)->additional_reserved_segments;
528 }
529
530 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
531 {
532 return FREE_I(sbi)->free_sections;
533 }
534
535 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
536 {
537 return DIRTY_I(sbi)->nr_dirty[PRE];
538 }
539
540 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
541 {
542 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
543 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
544 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
545 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
546 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
547 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
548 }
549
550 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
551 {
552 return SM_I(sbi)->ovp_segments;
553 }
554
555 static inline int reserved_sections(struct f2fs_sb_info *sbi)
556 {
557 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
558 }
559
560 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
561 unsigned int node_blocks, unsigned int dent_blocks)
562 {
563
564 unsigned segno, left_blocks;
565 int i;
566
567 /* check current node sections in the worst case. */
568 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
569 segno = CURSEG_I(sbi, i)->segno;
570 left_blocks = CAP_BLKS_PER_SEC(sbi) -
571 get_ckpt_valid_blocks(sbi, segno, true);
572 if (node_blocks > left_blocks)
573 return false;
574 }
575
576 /* check current data section for dentry blocks. */
577 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
578 left_blocks = CAP_BLKS_PER_SEC(sbi) -
579 get_ckpt_valid_blocks(sbi, segno, true);
580 if (dent_blocks > left_blocks)
581 return false;
582 return true;
583 }
584
585 /*
586 * calculate needed sections for dirty node/dentry
587 * and call has_curseg_enough_space
588 */
589 static inline void __get_secs_required(struct f2fs_sb_info *sbi,
590 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p)
591 {
592 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
593 get_pages(sbi, F2FS_DIRTY_DENTS) +
594 get_pages(sbi, F2FS_DIRTY_IMETA);
595 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
596 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi);
597 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi);
598 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi);
599 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi);
600
601 if (lower_p)
602 *lower_p = node_secs + dent_secs;
603 if (upper_p)
604 *upper_p = node_secs + dent_secs +
605 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
606 if (curseg_p)
607 *curseg_p = has_curseg_enough_space(sbi,
608 node_blocks, dent_blocks);
609 }
610
611 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
612 int freed, int needed)
613 {
614 unsigned int free_secs, lower_secs, upper_secs;
615 bool curseg_space;
616
617 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
618 return false;
619
620 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space);
621
622 free_secs = free_sections(sbi) + freed;
623 lower_secs += needed + reserved_sections(sbi);
624 upper_secs += needed + reserved_sections(sbi);
625
626 if (free_secs > upper_secs)
627 return false;
628 if (free_secs <= lower_secs)
629 return true;
630 return !curseg_space;
631 }
632
633 static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi,
634 int freed, int needed)
635 {
636 return !has_not_enough_free_secs(sbi, freed, needed);
637 }
638
639 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
640 {
641 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
642 return true;
643 if (likely(has_enough_free_secs(sbi, 0, 0)))
644 return true;
645 return false;
646 }
647
648 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
649 {
650 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
651 }
652
653 static inline int utilization(struct f2fs_sb_info *sbi)
654 {
655 return div_u64((u64)valid_user_blocks(sbi) * 100,
656 sbi->user_block_count);
657 }
658
659 /*
660 * Sometimes f2fs may be better to drop out-of-place update policy.
661 * And, users can control the policy through sysfs entries.
662 * There are five policies with triggering conditions as follows.
663 * F2FS_IPU_FORCE - all the time,
664 * F2FS_IPU_SSR - if SSR mode is activated,
665 * F2FS_IPU_UTIL - if FS utilization is over threashold,
666 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
667 * threashold,
668 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
669 * storages. IPU will be triggered only if the # of dirty
670 * pages over min_fsync_blocks. (=default option)
671 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
672 * F2FS_IPU_NOCACHE - disable IPU bio cache.
673 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has
674 * FI_OPU_WRITE flag.
675 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode)
676 */
677 #define DEF_MIN_IPU_UTIL 70
678 #define DEF_MIN_FSYNC_BLOCKS 8
679 #define DEF_MIN_HOT_BLOCKS 16
680
681 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
682
683 #define F2FS_IPU_DISABLE 0
684
685 /* Modification on enum should be synchronized with ipu_mode_names array */
686 enum {
687 F2FS_IPU_FORCE,
688 F2FS_IPU_SSR,
689 F2FS_IPU_UTIL,
690 F2FS_IPU_SSR_UTIL,
691 F2FS_IPU_FSYNC,
692 F2FS_IPU_ASYNC,
693 F2FS_IPU_NOCACHE,
694 F2FS_IPU_HONOR_OPU_WRITE,
695 F2FS_IPU_MAX,
696 };
697
698 static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi)
699 {
700 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE;
701 }
702
703 #define F2FS_IPU_POLICY(name) \
704 static inline bool IS_##name(struct f2fs_sb_info *sbi) \
705 { \
706 return SM_I(sbi)->ipu_policy & BIT(name); \
707 }
708
709 F2FS_IPU_POLICY(F2FS_IPU_FORCE);
710 F2FS_IPU_POLICY(F2FS_IPU_SSR);
711 F2FS_IPU_POLICY(F2FS_IPU_UTIL);
712 F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL);
713 F2FS_IPU_POLICY(F2FS_IPU_FSYNC);
714 F2FS_IPU_POLICY(F2FS_IPU_ASYNC);
715 F2FS_IPU_POLICY(F2FS_IPU_NOCACHE);
716 F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE);
717
718 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
719 int type)
720 {
721 struct curseg_info *curseg = CURSEG_I(sbi, type);
722 return curseg->segno;
723 }
724
725 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
726 int type)
727 {
728 struct curseg_info *curseg = CURSEG_I(sbi, type);
729 return curseg->alloc_type;
730 }
731
732 static inline bool valid_main_segno(struct f2fs_sb_info *sbi,
733 unsigned int segno)
734 {
735 return segno <= (MAIN_SEGS(sbi) - 1);
736 }
737
738 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
739 {
740 struct f2fs_sb_info *sbi = fio->sbi;
741
742 if (__is_valid_data_blkaddr(fio->old_blkaddr))
743 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
744 META_GENERIC : DATA_GENERIC);
745 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
746 META_GENERIC : DATA_GENERIC_ENHANCE);
747 }
748
749 /*
750 * Summary block is always treated as an invalid block
751 */
752 static inline int check_block_count(struct f2fs_sb_info *sbi,
753 int segno, struct f2fs_sit_entry *raw_sit)
754 {
755 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
756 int valid_blocks = 0;
757 int cur_pos = 0, next_pos;
758 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
759
760 /* check bitmap with valid block count */
761 do {
762 if (is_valid) {
763 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
764 usable_blks_per_seg,
765 cur_pos);
766 valid_blocks += next_pos - cur_pos;
767 } else
768 next_pos = find_next_bit_le(&raw_sit->valid_map,
769 usable_blks_per_seg,
770 cur_pos);
771 cur_pos = next_pos;
772 is_valid = !is_valid;
773 } while (cur_pos < usable_blks_per_seg);
774
775 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
776 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
777 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
778 set_sbi_flag(sbi, SBI_NEED_FSCK);
779 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
780 return -EFSCORRUPTED;
781 }
782
783 if (usable_blks_per_seg < BLKS_PER_SEG(sbi))
784 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
785 BLKS_PER_SEG(sbi),
786 usable_blks_per_seg) != BLKS_PER_SEG(sbi));
787
788 /* check segment usage, and check boundary of a given segment number */
789 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
790 || !valid_main_segno(sbi, segno))) {
791 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
792 GET_SIT_VBLOCKS(raw_sit), segno);
793 set_sbi_flag(sbi, SBI_NEED_FSCK);
794 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT);
795 return -EFSCORRUPTED;
796 }
797 return 0;
798 }
799
800 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
801 unsigned int start)
802 {
803 struct sit_info *sit_i = SIT_I(sbi);
804 unsigned int offset = SIT_BLOCK_OFFSET(start);
805 block_t blk_addr = sit_i->sit_base_addr + offset;
806
807 f2fs_bug_on(sbi, !valid_main_segno(sbi, start));
808
809 #ifdef CONFIG_F2FS_CHECK_FS
810 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
811 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
812 f2fs_bug_on(sbi, 1);
813 #endif
814
815 /* calculate sit block address */
816 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
817 blk_addr += sit_i->sit_blocks;
818
819 return blk_addr;
820 }
821
822 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
823 pgoff_t block_addr)
824 {
825 struct sit_info *sit_i = SIT_I(sbi);
826 block_addr -= sit_i->sit_base_addr;
827 if (block_addr < sit_i->sit_blocks)
828 block_addr += sit_i->sit_blocks;
829 else
830 block_addr -= sit_i->sit_blocks;
831
832 return block_addr + sit_i->sit_base_addr;
833 }
834
835 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
836 {
837 unsigned int block_off = SIT_BLOCK_OFFSET(start);
838
839 f2fs_change_bit(block_off, sit_i->sit_bitmap);
840 #ifdef CONFIG_F2FS_CHECK_FS
841 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
842 #endif
843 }
844
845 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
846 bool base_time)
847 {
848 struct sit_info *sit_i = SIT_I(sbi);
849 time64_t diff, now = ktime_get_boottime_seconds();
850
851 if (now >= sit_i->mounted_time)
852 return sit_i->elapsed_time + now - sit_i->mounted_time;
853
854 /* system time is set to the past */
855 if (!base_time) {
856 diff = sit_i->mounted_time - now;
857 if (sit_i->elapsed_time >= diff)
858 return sit_i->elapsed_time - diff;
859 return 0;
860 }
861 return sit_i->elapsed_time;
862 }
863
864 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
865 unsigned int ofs_in_node, unsigned char version)
866 {
867 sum->nid = cpu_to_le32(nid);
868 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
869 sum->version = version;
870 }
871
872 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
873 {
874 return __start_cp_addr(sbi) +
875 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
876 }
877
878 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
879 {
880 return __start_cp_addr(sbi) +
881 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
882 - (base + 1) + type;
883 }
884
885 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
886 {
887 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
888 return true;
889 return false;
890 }
891
892 /*
893 * It is very important to gather dirty pages and write at once, so that we can
894 * submit a big bio without interfering other data writes.
895 * By default, 512 pages for directory data,
896 * 512 pages (2MB) * 8 for nodes, and
897 * 256 pages * 8 for meta are set.
898 */
899 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
900 {
901 if (sbi->sb->s_bdi->wb.dirty_exceeded)
902 return 0;
903
904 if (type == DATA)
905 return BLKS_PER_SEG(sbi);
906 else if (type == NODE)
907 return SEGS_TO_BLKS(sbi, 8);
908 else if (type == META)
909 return 8 * BIO_MAX_VECS;
910 else
911 return 0;
912 }
913
914 /*
915 * When writing pages, it'd better align nr_to_write for segment size.
916 */
917 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
918 struct writeback_control *wbc)
919 {
920 long nr_to_write, desired;
921
922 if (wbc->sync_mode != WB_SYNC_NONE)
923 return 0;
924
925 nr_to_write = wbc->nr_to_write;
926 desired = BIO_MAX_VECS;
927 if (type == NODE)
928 desired <<= 1;
929
930 wbc->nr_to_write = desired;
931 return desired - nr_to_write;
932 }
933
934 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
935 {
936 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
937 bool wakeup = false;
938 int i;
939
940 if (force)
941 goto wake_up;
942
943 mutex_lock(&dcc->cmd_lock);
944 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
945 if (i + 1 < dcc->discard_granularity)
946 break;
947 if (!list_empty(&dcc->pend_list[i])) {
948 wakeup = true;
949 break;
950 }
951 }
952 mutex_unlock(&dcc->cmd_lock);
953 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
954 return;
955 wake_up:
956 dcc->discard_wake = true;
957 wake_up_interruptible_all(&dcc->discard_wait_queue);
958 }
959
960 static inline unsigned int first_zoned_segno(struct f2fs_sb_info *sbi)
961 {
962 int devi;
963
964 for (devi = 0; devi < sbi->s_ndevs; devi++)
965 if (bdev_is_zoned(FDEV(devi).bdev))
966 return GET_SEGNO(sbi, FDEV(devi).start_blk);
967 return 0;
968 }
969
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.