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