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
sanity_check_seg_type(struct f2fs_sb_info * sbi,unsigned short seg_type)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 (sbi)->segs_per_sec) || \
52 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
53 (sbi)->segs_per_sec) || \
54 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
55 (sbi)->segs_per_sec) || \
56 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
57 (sbi)->segs_per_sec) || \
58 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
59 (sbi)->segs_per_sec) || \
60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
61 (sbi)->segs_per_sec) || \
62 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \
63 (sbi)->segs_per_sec) || \
64 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \
65 (sbi)->segs_per_sec))
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) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
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 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
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 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
95 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
96 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 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 BLKS_PER_SEC(sbi) \
103 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
104 #define GET_SEC_FROM_SEG(sbi, segno) \
105 (((segno) == -1) ? -1: (segno) / (sbi)->segs_per_sec)
106 #define GET_SEG_FROM_SEC(sbi, secno) \
107 ((secno) * (sbi)->segs_per_sec)
108 #define GET_ZONE_FROM_SEC(sbi, secno) \
109 (((secno) == -1) ? -1: (secno) / (sbi)->secs_per_zone)
110 #define GET_ZONE_FROM_SEG(sbi, segno) \
111 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
112
113 #define GET_SUM_BLOCK(sbi, segno) \
114 ((sbi)->sm_info->ssa_blkaddr + (segno))
115
116 #define GET_SUM_TYPE(footer) ((footer)->entry_type)
117 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type))
118
119 #define SIT_ENTRY_OFFSET(sit_i, segno) \
120 ((segno) % (sit_i)->sents_per_block)
121 #define SIT_BLOCK_OFFSET(segno) \
122 ((segno) / SIT_ENTRY_PER_BLOCK)
123 #define START_SEGNO(segno) \
124 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
125 #define SIT_BLK_CNT(sbi) \
126 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK)
127 #define f2fs_bitmap_size(nr) \
128 (BITS_TO_LONGS(nr) * sizeof(unsigned long))
129
130 #define SECTOR_FROM_BLOCK(blk_addr) \
131 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
132 #define SECTOR_TO_BLOCK(sectors) \
133 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK)
134 #ifdef CONFIG_F2FS_GRADING_SSR
135 #define KBS_PER_SEGMENT 2048
136 #define SSR_MIN_BLKS_LIMIT (16 << 18) /* 16G */
137 #define SSR_CONTIG_DIRTY_NUMS 32 /* Dirty pages for LFS alloction in grading ssr. */
138 #define SSR_CONTIG_LARGE 256 /* Larege files */
139 #endif
140
141 enum {
142 SEQ_NONE,
143 SEQ_32BLKS,
144 SEQ_256BLKS
145 };
146 /*
147 * indicate a block allocation direction: RIGHT and LEFT.
148 * RIGHT means allocating new sections towards the end of volume.
149 * LEFT means the opposite direction.
150 */
151 enum {
152 ALLOC_RIGHT = 0,
153 ALLOC_LEFT
154 };
155
156 /*
157 * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
158 * LFS writes data sequentially with cleaning operations.
159 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
160 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
161 * fragmented segment which has similar aging degree.
162 */
163 enum {
164 LFS = 0,
165 SSR,
166 AT_SSR,
167 };
168
169 /*
170 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
171 * GC_CB is based on cost-benefit algorithm.
172 * GC_GREEDY is based on greedy algorithm.
173 * GC_AT is based on age-threshold algorithm.
174 */
175 enum {
176 GC_CB = 0,
177 GC_GREEDY,
178 GC_AT,
179 ALLOC_NEXT,
180 FLUSH_DEVICE,
181 MAX_GC_POLICY,
182 };
183
184 /*
185 * BG_GC means the background cleaning job.
186 * FG_GC means the on-demand cleaning job.
187 * FORCE_FG_GC means on-demand cleaning job in background.
188 */
189 enum {
190 BG_GC = 0,
191 FG_GC,
192 FORCE_FG_GC,
193 };
194
195 #ifdef CONFIG_F2FS_GRADING_SSR
196 enum {
197 GRADING_SSR_OFF = 0,
198 GRADING_SSR_ON
199 };
200 #endif
201
202 /* for a function parameter to select a victim segment */
203 struct victim_sel_policy {
204 int alloc_mode; /* LFS or SSR */
205 int gc_mode; /* GC_CB or GC_GREEDY */
206 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */
207 unsigned int max_search; /*
208 * maximum # of segments/sections
209 * to search
210 */
211 unsigned int offset; /* last scanned bitmap offset */
212 unsigned int ofs_unit; /* bitmap search unit */
213 unsigned int min_cost; /* minimum cost */
214 unsigned long long oldest_age; /* oldest age of segments having the same min cost */
215 unsigned int min_segno; /* segment # having min. cost */
216 unsigned long long age; /* mtime of GCed section*/
217 unsigned long long age_threshold;/* age threshold */
218 };
219
220 struct seg_entry {
221 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
222 unsigned int valid_blocks:10; /* # of valid blocks */
223 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
224 unsigned int padding:6; /* padding */
225 unsigned char *cur_valid_map; /* validity bitmap of blocks */
226 #ifdef CONFIG_F2FS_CHECK_FS
227 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
228 #endif
229 /*
230 * # of valid blocks and the validity bitmap stored in the last
231 * checkpoint pack. This information is used by the SSR mode.
232 */
233 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
234 unsigned char *discard_map;
235 unsigned long long mtime; /* modification time of the segment */
236 };
237
238 struct sec_entry {
239 unsigned int valid_blocks; /* # of valid blocks in a section */
240 };
241
242 struct segment_allocation {
243 void (*allocate_segment)(struct f2fs_sb_info *, int, bool, int);
244 };
245
246 #define MAX_SKIP_GC_COUNT 16
247
248 struct inmem_pages {
249 struct list_head list;
250 struct page *page;
251 block_t old_addr; /* for revoking when fail to commit */
252 };
253
254 struct sit_info {
255 const struct segment_allocation *s_ops;
256
257 block_t sit_base_addr; /* start block address of SIT area */
258 block_t sit_blocks; /* # of blocks used by SIT area */
259 block_t written_valid_blocks; /* # of valid blocks in main area */
260 char *bitmap; /* all bitmaps pointer */
261 char *sit_bitmap; /* SIT bitmap pointer */
262 #ifdef CONFIG_F2FS_CHECK_FS
263 char *sit_bitmap_mir; /* SIT bitmap mirror */
264
265 /* bitmap of segments to be ignored by GC in case of errors */
266 unsigned long *invalid_segmap;
267 #endif
268 unsigned int bitmap_size; /* SIT bitmap size */
269
270 unsigned long *tmp_map; /* bitmap for temporal use */
271 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
272 unsigned int dirty_sentries; /* # of dirty sentries */
273 unsigned int sents_per_block; /* # of SIT entries per block */
274 struct rw_semaphore sentry_lock; /* to protect SIT cache */
275 struct seg_entry *sentries; /* SIT segment-level cache */
276 struct sec_entry *sec_entries; /* SIT section-level cache */
277
278 /* for cost-benefit algorithm in cleaning procedure */
279 unsigned long long elapsed_time; /* elapsed time after mount */
280 unsigned long long mounted_time; /* mount time */
281 unsigned long long min_mtime; /* min. modification time */
282 unsigned long long max_mtime; /* max. modification time */
283 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */
284 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */
285
286 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
287 };
288
289 struct free_segmap_info {
290 unsigned int start_segno; /* start segment number logically */
291 unsigned int free_segments; /* # of free segments */
292 unsigned int free_sections; /* # of free sections */
293 spinlock_t segmap_lock; /* free segmap lock */
294 unsigned long *free_segmap; /* free segment bitmap */
295 unsigned long *free_secmap; /* free section bitmap */
296 };
297
298 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
299 enum dirty_type {
300 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
301 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
302 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
303 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
304 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
305 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
306 DIRTY, /* to count # of dirty segments */
307 PRE, /* to count # of entirely obsolete segments */
308 NR_DIRTY_TYPE
309 };
310
311 struct dirty_seglist_info {
312 const struct victim_selection *v_ops; /* victim selction operation */
313 unsigned long *dirty_segmap[NR_DIRTY_TYPE];
314 unsigned long *dirty_secmap;
315 struct mutex seglist_lock; /* lock for segment bitmaps */
316 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
317 unsigned long *victim_secmap; /* background GC victims */
318 };
319
320 /* victim selection function for cleaning and SSR */
321 struct victim_selection {
322 int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
323 int, int, char, unsigned long long);
324 };
325
326 /* for active log information */
327 struct curseg_info {
328 struct mutex curseg_mutex; /* lock for consistency */
329 struct f2fs_summary_block *sum_blk; /* cached summary block */
330 struct rw_semaphore journal_rwsem; /* protect journal area */
331 struct f2fs_journal *journal; /* cached journal info */
332 unsigned char alloc_type; /* current allocation type */
333 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */
334 unsigned int segno; /* current segment number */
335 unsigned short next_blkoff; /* next block offset to write */
336 unsigned int zone; /* current zone number */
337 unsigned int next_segno; /* preallocated segment */
338 bool inited; /* indicate inmem log is inited */
339 };
340
341 struct sit_entry_set {
342 struct list_head set_list; /* link with all sit sets */
343 unsigned int start_segno; /* start segno of sits in set */
344 unsigned int entry_cnt; /* the # of sit entries in set */
345 };
346
347 /*
348 * inline functions
349 */
CURSEG_I(struct f2fs_sb_info * sbi,int type)350 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
351 {
352 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
353 }
354
get_seg_entry(struct f2fs_sb_info * sbi,unsigned int segno)355 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
356 unsigned int segno)
357 {
358 struct sit_info *sit_i = SIT_I(sbi);
359 return &sit_i->sentries[segno];
360 }
361
get_sec_entry(struct f2fs_sb_info * sbi,unsigned int segno)362 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
363 unsigned int segno)
364 {
365 struct sit_info *sit_i = SIT_I(sbi);
366 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
367 }
368
get_valid_blocks(struct f2fs_sb_info * sbi,unsigned int segno,bool use_section)369 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
370 unsigned int segno, bool use_section)
371 {
372 /*
373 * In order to get # of valid blocks in a section instantly from many
374 * segments, f2fs manages two counting structures separately.
375 */
376 if (use_section && __is_large_section(sbi))
377 return get_sec_entry(sbi, segno)->valid_blocks;
378 else
379 return get_seg_entry(sbi, segno)->valid_blocks;
380 }
381
get_ckpt_valid_blocks(struct f2fs_sb_info * sbi,unsigned int segno,bool use_section)382 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi,
383 unsigned int segno, bool use_section)
384 {
385 if (use_section && __is_large_section(sbi)) {
386 unsigned int start_segno = START_SEGNO(segno);
387 unsigned int blocks = 0;
388 int i;
389
390 for (i = 0; i < sbi->segs_per_sec; i++, start_segno++) {
391 struct seg_entry *se = get_seg_entry(sbi, start_segno);
392
393 blocks += se->ckpt_valid_blocks;
394 }
395 return blocks;
396 }
397 return get_seg_entry(sbi, segno)->ckpt_valid_blocks;
398 }
399
seg_info_from_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)400 static inline void seg_info_from_raw_sit(struct seg_entry *se,
401 struct f2fs_sit_entry *rs)
402 {
403 se->valid_blocks = GET_SIT_VBLOCKS(rs);
404 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
405 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
406 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
407 #ifdef CONFIG_F2FS_CHECK_FS
408 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
409 #endif
410 se->type = GET_SIT_TYPE(rs);
411 se->mtime = le64_to_cpu(rs->mtime);
412 }
413
__seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)414 static inline void __seg_info_to_raw_sit(struct seg_entry *se,
415 struct f2fs_sit_entry *rs)
416 {
417 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
418 se->valid_blocks;
419 rs->vblocks = cpu_to_le16(raw_vblocks);
420 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
421 rs->mtime = cpu_to_le64(se->mtime);
422 }
423
seg_info_to_sit_page(struct f2fs_sb_info * sbi,struct page * page,unsigned int start)424 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi,
425 struct page *page, unsigned int start)
426 {
427 struct f2fs_sit_block *raw_sit;
428 struct seg_entry *se;
429 struct f2fs_sit_entry *rs;
430 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK,
431 (unsigned long)MAIN_SEGS(sbi));
432 int i;
433
434 raw_sit = (struct f2fs_sit_block *)page_address(page);
435 memset(raw_sit, 0, PAGE_SIZE);
436 for (i = 0; i < end - start; i++) {
437 rs = &raw_sit->entries[i];
438 se = get_seg_entry(sbi, start + i);
439 __seg_info_to_raw_sit(se, rs);
440 }
441 }
442
seg_info_to_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * rs)443 static inline void seg_info_to_raw_sit(struct seg_entry *se,
444 struct f2fs_sit_entry *rs)
445 {
446 __seg_info_to_raw_sit(se, rs);
447
448 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
449 se->ckpt_valid_blocks = se->valid_blocks;
450 }
451
find_next_inuse(struct free_segmap_info * free_i,unsigned int max,unsigned int segno)452 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
453 unsigned int max, unsigned int segno)
454 {
455 unsigned int ret;
456 spin_lock(&free_i->segmap_lock);
457 ret = find_next_bit(free_i->free_segmap, max, segno);
458 spin_unlock(&free_i->segmap_lock);
459 return ret;
460 }
461
__set_free(struct f2fs_sb_info * sbi,unsigned int segno)462 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
463 {
464 struct free_segmap_info *free_i = FREE_I(sbi);
465 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
466 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
467 unsigned int next;
468 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
469
470 spin_lock(&free_i->segmap_lock);
471 clear_bit(segno, free_i->free_segmap);
472 free_i->free_segments++;
473
474 next = find_next_bit(free_i->free_segmap,
475 start_segno + sbi->segs_per_sec, start_segno);
476 if (next >= start_segno + usable_segs) {
477 clear_bit(secno, free_i->free_secmap);
478 free_i->free_sections++;
479 }
480 spin_unlock(&free_i->segmap_lock);
481 }
482
__set_inuse(struct f2fs_sb_info * sbi,unsigned int segno)483 static inline void __set_inuse(struct f2fs_sb_info *sbi,
484 unsigned int segno)
485 {
486 struct free_segmap_info *free_i = FREE_I(sbi);
487 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
488
489 set_bit(segno, free_i->free_segmap);
490 free_i->free_segments--;
491 if (!test_and_set_bit(secno, free_i->free_secmap))
492 free_i->free_sections--;
493 }
494
__set_test_and_free(struct f2fs_sb_info * sbi,unsigned int segno,bool inmem)495 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
496 unsigned int segno, bool inmem)
497 {
498 struct free_segmap_info *free_i = FREE_I(sbi);
499 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
500 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
501 unsigned int next;
502 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno);
503
504 spin_lock(&free_i->segmap_lock);
505 if (test_and_clear_bit(segno, free_i->free_segmap)) {
506 free_i->free_segments++;
507
508 if (!inmem && IS_CURSEC(sbi, secno))
509 goto skip_free;
510 next = find_next_bit(free_i->free_segmap,
511 start_segno + sbi->segs_per_sec, start_segno);
512 if (next >= start_segno + usable_segs) {
513 if (test_and_clear_bit(secno, free_i->free_secmap))
514 free_i->free_sections++;
515 }
516 }
517 skip_free:
518 spin_unlock(&free_i->segmap_lock);
519 }
520
__set_test_and_inuse(struct f2fs_sb_info * sbi,unsigned int segno)521 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
522 unsigned int segno)
523 {
524 struct free_segmap_info *free_i = FREE_I(sbi);
525 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
526
527 spin_lock(&free_i->segmap_lock);
528 if (!test_and_set_bit(segno, free_i->free_segmap)) {
529 free_i->free_segments--;
530 if (!test_and_set_bit(secno, free_i->free_secmap))
531 free_i->free_sections--;
532 }
533 spin_unlock(&free_i->segmap_lock);
534 }
535
get_sit_bitmap(struct f2fs_sb_info * sbi,void * dst_addr)536 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
537 void *dst_addr)
538 {
539 struct sit_info *sit_i = SIT_I(sbi);
540
541 #ifdef CONFIG_F2FS_CHECK_FS
542 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
543 sit_i->bitmap_size))
544 f2fs_bug_on(sbi, 1);
545 #endif
546 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
547 }
548
written_block_count(struct f2fs_sb_info * sbi)549 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
550 {
551 return SIT_I(sbi)->written_valid_blocks;
552 }
553
free_segments(struct f2fs_sb_info * sbi)554 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
555 {
556 return FREE_I(sbi)->free_segments;
557 }
558
reserved_segments(struct f2fs_sb_info * sbi)559 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi)
560 {
561 return SM_I(sbi)->reserved_segments +
562 SM_I(sbi)->additional_reserved_segments;
563 }
564
free_sections(struct f2fs_sb_info * sbi)565 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
566 {
567 return FREE_I(sbi)->free_sections;
568 }
569
prefree_segments(struct f2fs_sb_info * sbi)570 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
571 {
572 return DIRTY_I(sbi)->nr_dirty[PRE];
573 }
574
dirty_segments(struct f2fs_sb_info * sbi)575 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
576 {
577 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
578 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
579 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
580 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
581 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
582 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
583 }
584
overprovision_segments(struct f2fs_sb_info * sbi)585 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
586 {
587 return SM_I(sbi)->ovp_segments;
588 }
589
reserved_sections(struct f2fs_sb_info * sbi)590 static inline int reserved_sections(struct f2fs_sb_info *sbi)
591 {
592 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi));
593 }
594
has_curseg_enough_space(struct f2fs_sb_info * sbi,unsigned int node_blocks,unsigned int dent_blocks)595 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi,
596 unsigned int node_blocks, unsigned int dent_blocks)
597 {
598
599 unsigned int segno, left_blocks;
600 int i;
601
602 /* check current node segment */
603 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) {
604 segno = CURSEG_I(sbi, i)->segno;
605 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
606 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
607
608 if (node_blocks > left_blocks)
609 return false;
610 }
611
612 /* check current data segment */
613 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno;
614 left_blocks = f2fs_usable_blks_in_seg(sbi, segno) -
615 get_seg_entry(sbi, segno)->ckpt_valid_blocks;
616 if (dent_blocks > left_blocks)
617 return false;
618 return true;
619 }
620
has_not_enough_free_secs(struct f2fs_sb_info * sbi,int freed,int needed)621 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
622 int freed, int needed)
623 {
624 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) +
625 get_pages(sbi, F2FS_DIRTY_DENTS) +
626 get_pages(sbi, F2FS_DIRTY_IMETA);
627 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS);
628 unsigned int node_secs = total_node_blocks / BLKS_PER_SEC(sbi);
629 unsigned int dent_secs = total_dent_blocks / BLKS_PER_SEC(sbi);
630 unsigned int node_blocks = total_node_blocks % BLKS_PER_SEC(sbi);
631 unsigned int dent_blocks = total_dent_blocks % BLKS_PER_SEC(sbi);
632 unsigned int free, need_lower, need_upper;
633
634 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
635 return false;
636
637 free = free_sections(sbi) + freed;
638 need_lower = node_secs + dent_secs + reserved_sections(sbi) + needed;
639 need_upper = need_lower + (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0);
640
641 if (free > need_upper)
642 return false;
643 else if (free <= need_lower)
644 return true;
645 return !has_curseg_enough_space(sbi, node_blocks, dent_blocks);
646 }
647
f2fs_is_checkpoint_ready(struct f2fs_sb_info * sbi)648 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi)
649 {
650 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
651 return true;
652 if (likely(!has_not_enough_free_secs(sbi, 0, 0)))
653 return true;
654 return false;
655 }
656
excess_prefree_segs(struct f2fs_sb_info * sbi)657 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
658 {
659 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
660 }
661
utilization(struct f2fs_sb_info * sbi)662 static inline int utilization(struct f2fs_sb_info *sbi)
663 {
664 return div_u64((u64)valid_user_blocks(sbi) * 100,
665 sbi->user_block_count);
666 }
667
668 /*
669 * Sometimes f2fs may be better to drop out-of-place update policy.
670 * And, users can control the policy through sysfs entries.
671 * There are five policies with triggering conditions as follows.
672 * F2FS_IPU_FORCE - all the time,
673 * F2FS_IPU_SSR - if SSR mode is activated,
674 * F2FS_IPU_UTIL - if FS utilization is over threashold,
675 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
676 * threashold,
677 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
678 * storages. IPU will be triggered only if the # of dirty
679 * pages over min_fsync_blocks. (=default option)
680 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests.
681 * F2FS_IPU_NOCACHE - disable IPU bio cache.
682 * F2FS_IPUT_DISABLE - disable IPU. (=default option in LFS mode)
683 */
684 #define DEF_MIN_IPU_UTIL 70
685 #define DEF_MIN_FSYNC_BLOCKS 8
686 #define DEF_MIN_HOT_BLOCKS 16
687
688 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */
689
690 enum {
691 F2FS_IPU_FORCE,
692 F2FS_IPU_SSR,
693 F2FS_IPU_UTIL,
694 F2FS_IPU_SSR_UTIL,
695 F2FS_IPU_FSYNC,
696 F2FS_IPU_ASYNC,
697 F2FS_IPU_NOCACHE,
698 };
699
curseg_segno(struct f2fs_sb_info * sbi,int type)700 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
701 int type)
702 {
703 struct curseg_info *curseg = CURSEG_I(sbi, type);
704 return curseg->segno;
705 }
706
curseg_alloc_type(struct f2fs_sb_info * sbi,int type)707 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
708 int type)
709 {
710 struct curseg_info *curseg = CURSEG_I(sbi, type);
711 return curseg->alloc_type;
712 }
713
curseg_blkoff(struct f2fs_sb_info * sbi,int type)714 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
715 {
716 struct curseg_info *curseg = CURSEG_I(sbi, type);
717 return curseg->next_blkoff;
718 }
719
check_seg_range(struct f2fs_sb_info * sbi,unsigned int segno)720 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
721 {
722 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
723 }
724
verify_fio_blkaddr(struct f2fs_io_info * fio)725 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio)
726 {
727 struct f2fs_sb_info *sbi = fio->sbi;
728
729 if (__is_valid_data_blkaddr(fio->old_blkaddr))
730 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ?
731 META_GENERIC : DATA_GENERIC);
732 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ?
733 META_GENERIC : DATA_GENERIC_ENHANCE);
734 }
735
736 /*
737 * Summary block is always treated as an invalid block
738 */
check_block_count(struct f2fs_sb_info * sbi,int segno,struct f2fs_sit_entry * raw_sit)739 static inline int check_block_count(struct f2fs_sb_info *sbi,
740 int segno, struct f2fs_sit_entry *raw_sit)
741 {
742 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
743 int valid_blocks = 0;
744 int cur_pos = 0, next_pos;
745 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno);
746
747 /* check bitmap with valid block count */
748 do {
749 if (is_valid) {
750 next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
751 usable_blks_per_seg,
752 cur_pos);
753 valid_blocks += next_pos - cur_pos;
754 } else
755 next_pos = find_next_bit_le(&raw_sit->valid_map,
756 usable_blks_per_seg,
757 cur_pos);
758 cur_pos = next_pos;
759 is_valid = !is_valid;
760 } while (cur_pos < usable_blks_per_seg);
761
762 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
763 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d",
764 GET_SIT_VBLOCKS(raw_sit), valid_blocks);
765 set_sbi_flag(sbi, SBI_NEED_FSCK);
766 return -EFSCORRUPTED;
767 }
768
769 if (usable_blks_per_seg < sbi->blocks_per_seg)
770 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map,
771 sbi->blocks_per_seg,
772 usable_blks_per_seg) != sbi->blocks_per_seg);
773
774 /* check segment usage, and check boundary of a given segment number */
775 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg
776 || segno > TOTAL_SEGS(sbi) - 1)) {
777 f2fs_err(sbi, "Wrong valid blocks %d or segno %u",
778 GET_SIT_VBLOCKS(raw_sit), segno);
779 set_sbi_flag(sbi, SBI_NEED_FSCK);
780 return -EFSCORRUPTED;
781 }
782 return 0;
783 }
784
current_sit_addr(struct f2fs_sb_info * sbi,unsigned int start)785 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
786 unsigned int start)
787 {
788 struct sit_info *sit_i = SIT_I(sbi);
789 unsigned int offset = SIT_BLOCK_OFFSET(start);
790 block_t blk_addr = sit_i->sit_base_addr + offset;
791
792 check_seg_range(sbi, start);
793
794 #ifdef CONFIG_F2FS_CHECK_FS
795 if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
796 f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
797 f2fs_bug_on(sbi, 1);
798 #endif
799
800 /* calculate sit block address */
801 if (f2fs_test_bit(offset, sit_i->sit_bitmap))
802 blk_addr += sit_i->sit_blocks;
803
804 return blk_addr;
805 }
806
next_sit_addr(struct f2fs_sb_info * sbi,pgoff_t block_addr)807 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
808 pgoff_t block_addr)
809 {
810 struct sit_info *sit_i = SIT_I(sbi);
811 block_addr -= sit_i->sit_base_addr;
812 if (block_addr < sit_i->sit_blocks)
813 block_addr += sit_i->sit_blocks;
814 else
815 block_addr -= sit_i->sit_blocks;
816
817 return block_addr + sit_i->sit_base_addr;
818 }
819
set_to_next_sit(struct sit_info * sit_i,unsigned int start)820 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
821 {
822 unsigned int block_off = SIT_BLOCK_OFFSET(start);
823
824 f2fs_change_bit(block_off, sit_i->sit_bitmap);
825 #ifdef CONFIG_F2FS_CHECK_FS
826 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
827 #endif
828 }
829
get_mtime(struct f2fs_sb_info * sbi,bool base_time)830 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi,
831 bool base_time)
832 {
833 struct sit_info *sit_i = SIT_I(sbi);
834 time64_t diff, now = ktime_get_boottime_seconds();
835
836 if (now >= sit_i->mounted_time)
837 return sit_i->elapsed_time + now - sit_i->mounted_time;
838
839 /* system time is set to the past */
840 if (!base_time) {
841 diff = sit_i->mounted_time - now;
842 if (sit_i->elapsed_time >= diff)
843 return sit_i->elapsed_time - diff;
844 return 0;
845 }
846 return sit_i->elapsed_time;
847 }
848
set_summary(struct f2fs_summary * sum,nid_t nid,unsigned int ofs_in_node,unsigned char version)849 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
850 unsigned int ofs_in_node, unsigned char version)
851 {
852 sum->nid = cpu_to_le32(nid);
853 sum->ofs_in_node = cpu_to_le16(ofs_in_node);
854 sum->version = version;
855 }
856
start_sum_block(struct f2fs_sb_info * sbi)857 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
858 {
859 return __start_cp_addr(sbi) +
860 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
861 }
862
sum_blk_addr(struct f2fs_sb_info * sbi,int base,int type)863 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
864 {
865 return __start_cp_addr(sbi) +
866 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
867 - (base + 1) + type;
868 }
869
sec_usage_check(struct f2fs_sb_info * sbi,unsigned int secno)870 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
871 {
872 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
873 return true;
874 return false;
875 }
876
877 /*
878 * It is very important to gather dirty pages and write at once, so that we can
879 * submit a big bio without interfering other data writes.
880 * By default, 512 pages for directory data,
881 * 512 pages (2MB) * 8 for nodes, and
882 * 256 pages * 8 for meta are set.
883 */
nr_pages_to_skip(struct f2fs_sb_info * sbi,int type)884 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
885 {
886 if (sbi->sb->s_bdi->wb.dirty_exceeded)
887 return 0;
888
889 if (type == DATA)
890 return sbi->blocks_per_seg;
891 else if (type == NODE)
892 return 8 * sbi->blocks_per_seg;
893 else if (type == META)
894 return 8 * BIO_MAX_PAGES;
895 else
896 return 0;
897 }
898
899 /*
900 * When writing pages, it'd better align nr_to_write for segment size.
901 */
nr_pages_to_write(struct f2fs_sb_info * sbi,int type,struct writeback_control * wbc)902 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
903 struct writeback_control *wbc)
904 {
905 long nr_to_write, desired;
906
907 if (wbc->sync_mode != WB_SYNC_NONE)
908 return 0;
909
910 nr_to_write = wbc->nr_to_write;
911 desired = BIO_MAX_PAGES;
912 if (type == NODE)
913 desired <<= 1;
914
915 wbc->nr_to_write = desired;
916 return desired - nr_to_write;
917 }
918
wake_up_discard_thread(struct f2fs_sb_info * sbi,bool force)919 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force)
920 {
921 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
922 bool wakeup = false;
923 int i;
924
925 if (force)
926 goto wake_up;
927
928 mutex_lock(&dcc->cmd_lock);
929 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
930 if (i + 1 < dcc->discard_granularity)
931 break;
932 if (!list_empty(&dcc->pend_list[i])) {
933 wakeup = true;
934 break;
935 }
936 }
937 mutex_unlock(&dcc->cmd_lock);
938 if (!wakeup || !is_idle(sbi, DISCARD_TIME))
939 return;
940 wake_up:
941 dcc->discard_wake = 1;
942 wake_up_interruptible_all(&dcc->discard_wait_queue);
943 }
944
945 #ifdef CONFIG_F2FS_GRADING_SSR
check_io_seq(int blks)946 static inline int check_io_seq(int blks)
947 {
948 if (blks >= SSR_CONTIG_LARGE)
949 return SEQ_256BLKS;
950 if (blks >= SSR_CONTIG_DIRTY_NUMS)
951 return SEQ_32BLKS;
952 return SEQ_NONE;
953 }
954 #endif
955