]>
Commit | Line | Data |
---|---|---|
0a8165d7 | 1 | /* |
39a53e0c JK |
2 | * fs/f2fs/segment.h |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
ac5d156c | 11 | #include <linux/blkdev.h> |
66114cad | 12 | #include <linux/backing-dev.h> |
ac5d156c | 13 | |
39a53e0c JK |
14 | /* constant macro */ |
15 | #define NULL_SEGNO ((unsigned int)(~0)) | |
5ec4e49f | 16 | #define NULL_SECNO ((unsigned int)(~0)) |
39a53e0c | 17 | |
58c41035 | 18 | #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */ |
44a83499 | 19 | #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */ |
81eb8d6e | 20 | |
2040fce8 JK |
21 | #define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */ |
22 | ||
6224da87 | 23 | /* L: Logical segment # in volume, R: Relative segment # in main area */ |
68afcf2d TK |
24 | #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno) |
25 | #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno) | |
39a53e0c | 26 | |
68afcf2d TK |
27 | #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA) |
28 | #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE) | |
39a53e0c | 29 | |
5c773ba3 | 30 | #define IS_CURSEG(sbi, seg) \ |
68afcf2d TK |
31 | (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \ |
32 | ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \ | |
33 | ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \ | |
34 | ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \ | |
35 | ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \ | |
36 | ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno)) | |
39a53e0c JK |
37 | |
38 | #define IS_CURSEC(sbi, secno) \ | |
68afcf2d TK |
39 | (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \ |
40 | (sbi)->segs_per_sec) || \ | |
41 | ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \ | |
42 | (sbi)->segs_per_sec) || \ | |
43 | ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \ | |
44 | (sbi)->segs_per_sec) || \ | |
45 | ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \ | |
46 | (sbi)->segs_per_sec) || \ | |
47 | ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \ | |
48 | (sbi)->segs_per_sec) || \ | |
49 | ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \ | |
50 | (sbi)->segs_per_sec)) \ | |
39a53e0c | 51 | |
7cd8558b JK |
52 | #define MAIN_BLKADDR(sbi) (SM_I(sbi)->main_blkaddr) |
53 | #define SEG0_BLKADDR(sbi) (SM_I(sbi)->seg0_blkaddr) | |
54 | ||
55 | #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments) | |
68afcf2d | 56 | #define MAIN_SECS(sbi) ((sbi)->total_sections) |
7cd8558b JK |
57 | |
58 | #define TOTAL_SEGS(sbi) (SM_I(sbi)->segment_count) | |
68afcf2d | 59 | #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg) |
7cd8558b JK |
60 | |
61 | #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi)) | |
68afcf2d TK |
62 | #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \ |
63 | (sbi)->log_blocks_per_seg)) | |
7cd8558b JK |
64 | |
65 | #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \ | |
68afcf2d | 66 | (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg)) |
7cd8558b | 67 | |
39a53e0c | 68 | #define NEXT_FREE_BLKADDR(sbi, curseg) \ |
68afcf2d | 69 | (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff) |
39a53e0c | 70 | |
7cd8558b | 71 | #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi)) |
39a53e0c | 72 | #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ |
68afcf2d | 73 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg) |
491c0854 | 74 | #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ |
68afcf2d | 75 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1)) |
491c0854 | 76 | |
39a53e0c | 77 | #define GET_SEGNO(sbi, blk_addr) \ |
68afcf2d | 78 | ((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ? \ |
39a53e0c JK |
79 | NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ |
80 | GET_SEGNO_FROM_SEG0(sbi, blk_addr))) | |
4ddb1a4d JK |
81 | #define BLKS_PER_SEC(sbi) \ |
82 | ((sbi)->segs_per_sec * (sbi)->blocks_per_seg) | |
83 | #define GET_SEC_FROM_SEG(sbi, segno) \ | |
68afcf2d | 84 | ((segno) / (sbi)->segs_per_sec) |
4ddb1a4d | 85 | #define GET_SEG_FROM_SEC(sbi, secno) \ |
63fcf8e8 | 86 | ((secno) * (sbi)->segs_per_sec) |
4ddb1a4d JK |
87 | #define GET_ZONE_FROM_SEC(sbi, secno) \ |
88 | ((secno) / (sbi)->secs_per_zone) | |
89 | #define GET_ZONE_FROM_SEG(sbi, segno) \ | |
90 | GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) | |
39a53e0c JK |
91 | |
92 | #define GET_SUM_BLOCK(sbi, segno) \ | |
68afcf2d | 93 | ((sbi)->sm_info->ssa_blkaddr + (segno)) |
39a53e0c JK |
94 | |
95 | #define GET_SUM_TYPE(footer) ((footer)->entry_type) | |
68afcf2d | 96 | #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type)) |
39a53e0c JK |
97 | |
98 | #define SIT_ENTRY_OFFSET(sit_i, segno) \ | |
68afcf2d | 99 | ((segno) % (sit_i)->sents_per_block) |
d3a14afd | 100 | #define SIT_BLOCK_OFFSET(segno) \ |
68afcf2d | 101 | ((segno) / SIT_ENTRY_PER_BLOCK) |
d3a14afd CY |
102 | #define START_SEGNO(segno) \ |
103 | (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK) | |
74de593a | 104 | #define SIT_BLK_CNT(sbi) \ |
7cd8558b | 105 | ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK) |
39a53e0c JK |
106 | #define f2fs_bitmap_size(nr) \ |
107 | (BITS_TO_LONGS(nr) * sizeof(unsigned long)) | |
39a53e0c | 108 | |
55cf9cb6 CY |
109 | #define SECTOR_FROM_BLOCK(blk_addr) \ |
110 | (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK) | |
111 | #define SECTOR_TO_BLOCK(sectors) \ | |
68afcf2d | 112 | ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK) |
3cd8a239 | 113 | |
39a53e0c JK |
114 | /* |
115 | * indicate a block allocation direction: RIGHT and LEFT. | |
116 | * RIGHT means allocating new sections towards the end of volume. | |
117 | * LEFT means the opposite direction. | |
118 | */ | |
119 | enum { | |
120 | ALLOC_RIGHT = 0, | |
121 | ALLOC_LEFT | |
122 | }; | |
123 | ||
124 | /* | |
125 | * In the victim_sel_policy->alloc_mode, there are two block allocation modes. | |
126 | * LFS writes data sequentially with cleaning operations. | |
127 | * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. | |
128 | */ | |
129 | enum { | |
130 | LFS = 0, | |
131 | SSR | |
132 | }; | |
133 | ||
134 | /* | |
135 | * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes. | |
136 | * GC_CB is based on cost-benefit algorithm. | |
137 | * GC_GREEDY is based on greedy algorithm. | |
138 | */ | |
139 | enum { | |
140 | GC_CB = 0, | |
e066b83c JK |
141 | GC_GREEDY, |
142 | ALLOC_NEXT, | |
143 | FLUSH_DEVICE, | |
144 | MAX_GC_POLICY, | |
39a53e0c JK |
145 | }; |
146 | ||
147 | /* | |
148 | * BG_GC means the background cleaning job. | |
149 | * FG_GC means the on-demand cleaning job. | |
6aefd93b | 150 | * FORCE_FG_GC means on-demand cleaning job in background. |
39a53e0c JK |
151 | */ |
152 | enum { | |
153 | BG_GC = 0, | |
6aefd93b JK |
154 | FG_GC, |
155 | FORCE_FG_GC, | |
39a53e0c JK |
156 | }; |
157 | ||
158 | /* for a function parameter to select a victim segment */ | |
159 | struct victim_sel_policy { | |
160 | int alloc_mode; /* LFS or SSR */ | |
161 | int gc_mode; /* GC_CB or GC_GREEDY */ | |
162 | unsigned long *dirty_segmap; /* dirty segment bitmap */ | |
a26b7c8a | 163 | unsigned int max_search; /* maximum # of segments to search */ |
39a53e0c JK |
164 | unsigned int offset; /* last scanned bitmap offset */ |
165 | unsigned int ofs_unit; /* bitmap search unit */ | |
166 | unsigned int min_cost; /* minimum cost */ | |
167 | unsigned int min_segno; /* segment # having min. cost */ | |
168 | }; | |
169 | ||
170 | struct seg_entry { | |
f51b4ce6 CY |
171 | unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ |
172 | unsigned int valid_blocks:10; /* # of valid blocks */ | |
173 | unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ | |
174 | unsigned int padding:6; /* padding */ | |
39a53e0c | 175 | unsigned char *cur_valid_map; /* validity bitmap of blocks */ |
355e7891 CY |
176 | #ifdef CONFIG_F2FS_CHECK_FS |
177 | unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ | |
178 | #endif | |
39a53e0c JK |
179 | /* |
180 | * # of valid blocks and the validity bitmap stored in the the last | |
181 | * checkpoint pack. This information is used by the SSR mode. | |
182 | */ | |
f51b4ce6 | 183 | unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ |
a66cdd98 | 184 | unsigned char *discard_map; |
39a53e0c JK |
185 | unsigned long long mtime; /* modification time of the segment */ |
186 | }; | |
187 | ||
188 | struct sec_entry { | |
189 | unsigned int valid_blocks; /* # of valid blocks in a section */ | |
190 | }; | |
191 | ||
192 | struct segment_allocation { | |
193 | void (*allocate_segment)(struct f2fs_sb_info *, int, bool); | |
194 | }; | |
195 | ||
decd36b6 CY |
196 | /* |
197 | * this value is set in page as a private data which indicate that | |
198 | * the page is atomically written, and it is in inmem_pages list. | |
199 | */ | |
d48dfc20 | 200 | #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1) |
0a595eba | 201 | #define DUMMY_WRITTEN_PAGE ((unsigned long)-2) |
decd36b6 CY |
202 | |
203 | #define IS_ATOMIC_WRITTEN_PAGE(page) \ | |
204 | (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE) | |
0a595eba JK |
205 | #define IS_DUMMY_WRITTEN_PAGE(page) \ |
206 | (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE) | |
decd36b6 | 207 | |
88b88a66 JK |
208 | struct inmem_pages { |
209 | struct list_head list; | |
210 | struct page *page; | |
28bc106b | 211 | block_t old_addr; /* for revoking when fail to commit */ |
88b88a66 JK |
212 | }; |
213 | ||
39a53e0c JK |
214 | struct sit_info { |
215 | const struct segment_allocation *s_ops; | |
216 | ||
217 | block_t sit_base_addr; /* start block address of SIT area */ | |
218 | block_t sit_blocks; /* # of blocks used by SIT area */ | |
219 | block_t written_valid_blocks; /* # of valid blocks in main area */ | |
220 | char *sit_bitmap; /* SIT bitmap pointer */ | |
ae27d62e CY |
221 | #ifdef CONFIG_F2FS_CHECK_FS |
222 | char *sit_bitmap_mir; /* SIT bitmap mirror */ | |
223 | #endif | |
39a53e0c JK |
224 | unsigned int bitmap_size; /* SIT bitmap size */ |
225 | ||
60a3b782 | 226 | unsigned long *tmp_map; /* bitmap for temporal use */ |
39a53e0c JK |
227 | unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ |
228 | unsigned int dirty_sentries; /* # of dirty sentries */ | |
229 | unsigned int sents_per_block; /* # of SIT entries per block */ | |
230 | struct mutex sentry_lock; /* to protect SIT cache */ | |
231 | struct seg_entry *sentries; /* SIT segment-level cache */ | |
232 | struct sec_entry *sec_entries; /* SIT section-level cache */ | |
233 | ||
234 | /* for cost-benefit algorithm in cleaning procedure */ | |
235 | unsigned long long elapsed_time; /* elapsed time after mount */ | |
236 | unsigned long long mounted_time; /* mount time */ | |
237 | unsigned long long min_mtime; /* min. modification time */ | |
238 | unsigned long long max_mtime; /* max. modification time */ | |
e066b83c JK |
239 | |
240 | unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */ | |
39a53e0c JK |
241 | }; |
242 | ||
243 | struct free_segmap_info { | |
244 | unsigned int start_segno; /* start segment number logically */ | |
245 | unsigned int free_segments; /* # of free segments */ | |
246 | unsigned int free_sections; /* # of free sections */ | |
1a118ccf | 247 | spinlock_t segmap_lock; /* free segmap lock */ |
39a53e0c JK |
248 | unsigned long *free_segmap; /* free segment bitmap */ |
249 | unsigned long *free_secmap; /* free section bitmap */ | |
250 | }; | |
251 | ||
252 | /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ | |
253 | enum dirty_type { | |
254 | DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ | |
255 | DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ | |
256 | DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ | |
257 | DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ | |
258 | DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ | |
259 | DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ | |
260 | DIRTY, /* to count # of dirty segments */ | |
261 | PRE, /* to count # of entirely obsolete segments */ | |
262 | NR_DIRTY_TYPE | |
263 | }; | |
264 | ||
265 | struct dirty_seglist_info { | |
266 | const struct victim_selection *v_ops; /* victim selction operation */ | |
267 | unsigned long *dirty_segmap[NR_DIRTY_TYPE]; | |
268 | struct mutex seglist_lock; /* lock for segment bitmaps */ | |
269 | int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ | |
5ec4e49f | 270 | unsigned long *victim_secmap; /* background GC victims */ |
39a53e0c JK |
271 | }; |
272 | ||
273 | /* victim selection function for cleaning and SSR */ | |
274 | struct victim_selection { | |
275 | int (*get_victim)(struct f2fs_sb_info *, unsigned int *, | |
276 | int, int, char); | |
277 | }; | |
278 | ||
279 | /* for active log information */ | |
280 | struct curseg_info { | |
281 | struct mutex curseg_mutex; /* lock for consistency */ | |
282 | struct f2fs_summary_block *sum_blk; /* cached summary block */ | |
b7ad7512 CY |
283 | struct rw_semaphore journal_rwsem; /* protect journal area */ |
284 | struct f2fs_journal *journal; /* cached journal info */ | |
39a53e0c JK |
285 | unsigned char alloc_type; /* current allocation type */ |
286 | unsigned int segno; /* current segment number */ | |
287 | unsigned short next_blkoff; /* next block offset to write */ | |
288 | unsigned int zone; /* current zone number */ | |
289 | unsigned int next_segno; /* preallocated segment */ | |
290 | }; | |
291 | ||
184a5cd2 CY |
292 | struct sit_entry_set { |
293 | struct list_head set_list; /* link with all sit sets */ | |
294 | unsigned int start_segno; /* start segno of sits in set */ | |
295 | unsigned int entry_cnt; /* the # of sit entries in set */ | |
296 | }; | |
297 | ||
39a53e0c JK |
298 | /* |
299 | * inline functions | |
300 | */ | |
301 | static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) | |
302 | { | |
303 | return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); | |
304 | } | |
305 | ||
306 | static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, | |
307 | unsigned int segno) | |
308 | { | |
309 | struct sit_info *sit_i = SIT_I(sbi); | |
310 | return &sit_i->sentries[segno]; | |
311 | } | |
312 | ||
313 | static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, | |
314 | unsigned int segno) | |
315 | { | |
316 | struct sit_info *sit_i = SIT_I(sbi); | |
4ddb1a4d | 317 | return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; |
39a53e0c JK |
318 | } |
319 | ||
320 | static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, | |
302bd348 | 321 | unsigned int segno, bool use_section) |
39a53e0c JK |
322 | { |
323 | /* | |
324 | * In order to get # of valid blocks in a section instantly from many | |
325 | * segments, f2fs manages two counting structures separately. | |
326 | */ | |
302bd348 | 327 | if (use_section && sbi->segs_per_sec > 1) |
39a53e0c JK |
328 | return get_sec_entry(sbi, segno)->valid_blocks; |
329 | else | |
330 | return get_seg_entry(sbi, segno)->valid_blocks; | |
331 | } | |
332 | ||
333 | static inline void seg_info_from_raw_sit(struct seg_entry *se, | |
334 | struct f2fs_sit_entry *rs) | |
335 | { | |
336 | se->valid_blocks = GET_SIT_VBLOCKS(rs); | |
337 | se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); | |
338 | memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
339 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
355e7891 CY |
340 | #ifdef CONFIG_F2FS_CHECK_FS |
341 | memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
342 | #endif | |
39a53e0c JK |
343 | se->type = GET_SIT_TYPE(rs); |
344 | se->mtime = le64_to_cpu(rs->mtime); | |
345 | } | |
346 | ||
347 | static inline void seg_info_to_raw_sit(struct seg_entry *se, | |
348 | struct f2fs_sit_entry *rs) | |
349 | { | |
350 | unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | | |
351 | se->valid_blocks; | |
352 | rs->vblocks = cpu_to_le16(raw_vblocks); | |
353 | memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); | |
354 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
355 | se->ckpt_valid_blocks = se->valid_blocks; | |
356 | rs->mtime = cpu_to_le64(se->mtime); | |
357 | } | |
358 | ||
359 | static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, | |
360 | unsigned int max, unsigned int segno) | |
361 | { | |
362 | unsigned int ret; | |
1a118ccf | 363 | spin_lock(&free_i->segmap_lock); |
39a53e0c | 364 | ret = find_next_bit(free_i->free_segmap, max, segno); |
1a118ccf | 365 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
366 | return ret; |
367 | } | |
368 | ||
369 | static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) | |
370 | { | |
371 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
372 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
373 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
374 | unsigned int next; |
375 | ||
1a118ccf | 376 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
377 | clear_bit(segno, free_i->free_segmap); |
378 | free_i->free_segments++; | |
379 | ||
7fd97019 WL |
380 | next = find_next_bit(free_i->free_segmap, |
381 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
382 | if (next >= start_segno + sbi->segs_per_sec) { |
383 | clear_bit(secno, free_i->free_secmap); | |
384 | free_i->free_sections++; | |
385 | } | |
1a118ccf | 386 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
387 | } |
388 | ||
389 | static inline void __set_inuse(struct f2fs_sb_info *sbi, | |
390 | unsigned int segno) | |
391 | { | |
392 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
393 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
394 | ||
39a53e0c JK |
395 | set_bit(segno, free_i->free_segmap); |
396 | free_i->free_segments--; | |
397 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
398 | free_i->free_sections--; | |
399 | } | |
400 | ||
401 | static inline void __set_test_and_free(struct f2fs_sb_info *sbi, | |
402 | unsigned int segno) | |
403 | { | |
404 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
405 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
406 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
407 | unsigned int next; |
408 | ||
1a118ccf | 409 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
410 | if (test_and_clear_bit(segno, free_i->free_segmap)) { |
411 | free_i->free_segments++; | |
412 | ||
f1121ab0 CY |
413 | next = find_next_bit(free_i->free_segmap, |
414 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
415 | if (next >= start_segno + sbi->segs_per_sec) { |
416 | if (test_and_clear_bit(secno, free_i->free_secmap)) | |
417 | free_i->free_sections++; | |
418 | } | |
419 | } | |
1a118ccf | 420 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
421 | } |
422 | ||
423 | static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, | |
424 | unsigned int segno) | |
425 | { | |
426 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
427 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
428 | ||
1a118ccf | 429 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
430 | if (!test_and_set_bit(segno, free_i->free_segmap)) { |
431 | free_i->free_segments--; | |
432 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
433 | free_i->free_sections--; | |
434 | } | |
1a118ccf | 435 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
436 | } |
437 | ||
438 | static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, | |
439 | void *dst_addr) | |
440 | { | |
441 | struct sit_info *sit_i = SIT_I(sbi); | |
ae27d62e CY |
442 | |
443 | #ifdef CONFIG_F2FS_CHECK_FS | |
444 | if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, | |
445 | sit_i->bitmap_size)) | |
446 | f2fs_bug_on(sbi, 1); | |
447 | #endif | |
39a53e0c JK |
448 | memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); |
449 | } | |
450 | ||
451 | static inline block_t written_block_count(struct f2fs_sb_info *sbi) | |
452 | { | |
8b8343fa | 453 | return SIT_I(sbi)->written_valid_blocks; |
39a53e0c JK |
454 | } |
455 | ||
456 | static inline unsigned int free_segments(struct f2fs_sb_info *sbi) | |
457 | { | |
8b8343fa | 458 | return FREE_I(sbi)->free_segments; |
39a53e0c JK |
459 | } |
460 | ||
461 | static inline int reserved_segments(struct f2fs_sb_info *sbi) | |
462 | { | |
463 | return SM_I(sbi)->reserved_segments; | |
464 | } | |
465 | ||
466 | static inline unsigned int free_sections(struct f2fs_sb_info *sbi) | |
467 | { | |
8b8343fa | 468 | return FREE_I(sbi)->free_sections; |
39a53e0c JK |
469 | } |
470 | ||
471 | static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) | |
472 | { | |
473 | return DIRTY_I(sbi)->nr_dirty[PRE]; | |
474 | } | |
475 | ||
476 | static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) | |
477 | { | |
478 | return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + | |
479 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + | |
480 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + | |
481 | DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + | |
482 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + | |
483 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; | |
484 | } | |
485 | ||
486 | static inline int overprovision_segments(struct f2fs_sb_info *sbi) | |
487 | { | |
488 | return SM_I(sbi)->ovp_segments; | |
489 | } | |
490 | ||
491 | static inline int overprovision_sections(struct f2fs_sb_info *sbi) | |
492 | { | |
4ddb1a4d | 493 | return GET_SEC_FROM_SEG(sbi, (unsigned int)overprovision_segments(sbi)); |
39a53e0c JK |
494 | } |
495 | ||
496 | static inline int reserved_sections(struct f2fs_sb_info *sbi) | |
497 | { | |
4ddb1a4d | 498 | return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi)); |
39a53e0c JK |
499 | } |
500 | ||
501 | static inline bool need_SSR(struct f2fs_sb_info *sbi) | |
502 | { | |
95dd8973 JK |
503 | int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); |
504 | int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); | |
b9610bdf | 505 | int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA); |
36abef4e JK |
506 | |
507 | if (test_opt(sbi, LFS)) | |
508 | return false; | |
509 | ||
b9610bdf | 510 | return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs + |
796dbbfe | 511 | 2 * reserved_sections(sbi)); |
39a53e0c JK |
512 | } |
513 | ||
7f3037a5 JK |
514 | static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, |
515 | int freed, int needed) | |
39a53e0c | 516 | { |
5ac206cf NJ |
517 | int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); |
518 | int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); | |
b9610bdf | 519 | int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA); |
0f18b462 | 520 | |
caf0047e | 521 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
029cd28c JK |
522 | return false; |
523 | ||
7f3037a5 | 524 | return (free_sections(sbi) + freed) <= |
b9610bdf JK |
525 | (node_secs + 2 * dent_secs + imeta_secs + |
526 | reserved_sections(sbi) + needed); | |
39a53e0c JK |
527 | } |
528 | ||
81eb8d6e JK |
529 | static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) |
530 | { | |
6c311ec6 | 531 | return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; |
81eb8d6e JK |
532 | } |
533 | ||
39a53e0c JK |
534 | static inline int utilization(struct f2fs_sb_info *sbi) |
535 | { | |
6c311ec6 CF |
536 | return div_u64((u64)valid_user_blocks(sbi) * 100, |
537 | sbi->user_block_count); | |
39a53e0c JK |
538 | } |
539 | ||
540 | /* | |
541 | * Sometimes f2fs may be better to drop out-of-place update policy. | |
216fbd64 JK |
542 | * And, users can control the policy through sysfs entries. |
543 | * There are five policies with triggering conditions as follows. | |
544 | * F2FS_IPU_FORCE - all the time, | |
545 | * F2FS_IPU_SSR - if SSR mode is activated, | |
546 | * F2FS_IPU_UTIL - if FS utilization is over threashold, | |
547 | * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over | |
548 | * threashold, | |
c1ce1b02 JK |
549 | * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash |
550 | * storages. IPU will be triggered only if the # of dirty | |
551 | * pages over min_fsync_blocks. | |
216fbd64 | 552 | * F2FS_IPUT_DISABLE - disable IPU. (=default option) |
39a53e0c | 553 | */ |
216fbd64 | 554 | #define DEF_MIN_IPU_UTIL 70 |
c1ce1b02 | 555 | #define DEF_MIN_FSYNC_BLOCKS 8 |
ef095d19 | 556 | #define DEF_MIN_HOT_BLOCKS 16 |
216fbd64 JK |
557 | |
558 | enum { | |
559 | F2FS_IPU_FORCE, | |
560 | F2FS_IPU_SSR, | |
561 | F2FS_IPU_UTIL, | |
562 | F2FS_IPU_SSR_UTIL, | |
c1ce1b02 | 563 | F2FS_IPU_FSYNC, |
04485987 | 564 | F2FS_IPU_ASYNC, |
216fbd64 JK |
565 | }; |
566 | ||
7eab0c0d | 567 | static inline bool need_inplace_update_policy(struct inode *inode, |
04485987 | 568 | struct f2fs_io_info *fio) |
39a53e0c | 569 | { |
4081363f | 570 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9b5f136f | 571 | unsigned int policy = SM_I(sbi)->ipu_policy; |
216fbd64 | 572 | |
36abef4e JK |
573 | if (test_opt(sbi, LFS)) |
574 | return false; | |
575 | ||
9b5f136f | 576 | if (policy & (0x1 << F2FS_IPU_FORCE)) |
39a53e0c | 577 | return true; |
9b5f136f JK |
578 | if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi)) |
579 | return true; | |
580 | if (policy & (0x1 << F2FS_IPU_UTIL) && | |
581 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
582 | return true; | |
583 | if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) && | |
584 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
585 | return true; | |
586 | ||
04485987 HP |
587 | /* |
588 | * IPU for rewrite async pages | |
589 | */ | |
590 | if (policy & (0x1 << F2FS_IPU_ASYNC) && | |
591 | fio && fio->op == REQ_OP_WRITE && | |
4086d3f6 HP |
592 | !(fio->op_flags & REQ_SYNC) && |
593 | !f2fs_encrypted_inode(inode)) | |
04485987 HP |
594 | return true; |
595 | ||
9b5f136f JK |
596 | /* this is only set during fdatasync */ |
597 | if (policy & (0x1 << F2FS_IPU_FSYNC) && | |
91942321 | 598 | is_inode_flag_set(inode, FI_NEED_IPU)) |
9b5f136f JK |
599 | return true; |
600 | ||
39a53e0c JK |
601 | return false; |
602 | } | |
603 | ||
604 | static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, | |
605 | int type) | |
606 | { | |
607 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
608 | return curseg->segno; | |
609 | } | |
610 | ||
611 | static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, | |
612 | int type) | |
613 | { | |
614 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
615 | return curseg->alloc_type; | |
616 | } | |
617 | ||
618 | static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type) | |
619 | { | |
620 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
621 | return curseg->next_blkoff; | |
622 | } | |
623 | ||
624 | static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) | |
625 | { | |
7a04f64d | 626 | f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1); |
39a53e0c JK |
627 | } |
628 | ||
39a53e0c JK |
629 | static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr) |
630 | { | |
bb413d6a YH |
631 | BUG_ON(blk_addr < SEG0_BLKADDR(sbi) |
632 | || blk_addr >= MAX_BLKADDR(sbi)); | |
39a53e0c JK |
633 | } |
634 | ||
635 | /* | |
e1c42045 | 636 | * Summary block is always treated as an invalid block |
39a53e0c JK |
637 | */ |
638 | static inline void check_block_count(struct f2fs_sb_info *sbi, | |
639 | int segno, struct f2fs_sit_entry *raw_sit) | |
640 | { | |
4c278394 | 641 | #ifdef CONFIG_F2FS_CHECK_FS |
44c60bf2 | 642 | bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; |
39a53e0c | 643 | int valid_blocks = 0; |
44c60bf2 | 644 | int cur_pos = 0, next_pos; |
39a53e0c | 645 | |
39a53e0c | 646 | /* check bitmap with valid block count */ |
44c60bf2 CY |
647 | do { |
648 | if (is_valid) { | |
649 | next_pos = find_next_zero_bit_le(&raw_sit->valid_map, | |
650 | sbi->blocks_per_seg, | |
651 | cur_pos); | |
652 | valid_blocks += next_pos - cur_pos; | |
653 | } else | |
654 | next_pos = find_next_bit_le(&raw_sit->valid_map, | |
655 | sbi->blocks_per_seg, | |
656 | cur_pos); | |
657 | cur_pos = next_pos; | |
658 | is_valid = !is_valid; | |
659 | } while (cur_pos < sbi->blocks_per_seg); | |
39a53e0c | 660 | BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks); |
5d56b671 | 661 | #endif |
4c278394 JK |
662 | /* check segment usage, and check boundary of a given segment number */ |
663 | f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg | |
664 | || segno > TOTAL_SEGS(sbi) - 1); | |
7a04f64d | 665 | } |
39a53e0c JK |
666 | |
667 | static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, | |
668 | unsigned int start) | |
669 | { | |
670 | struct sit_info *sit_i = SIT_I(sbi); | |
d3a14afd | 671 | unsigned int offset = SIT_BLOCK_OFFSET(start); |
39a53e0c JK |
672 | block_t blk_addr = sit_i->sit_base_addr + offset; |
673 | ||
674 | check_seg_range(sbi, start); | |
675 | ||
ae27d62e CY |
676 | #ifdef CONFIG_F2FS_CHECK_FS |
677 | if (f2fs_test_bit(offset, sit_i->sit_bitmap) != | |
678 | f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) | |
679 | f2fs_bug_on(sbi, 1); | |
680 | #endif | |
681 | ||
39a53e0c JK |
682 | /* calculate sit block address */ |
683 | if (f2fs_test_bit(offset, sit_i->sit_bitmap)) | |
684 | blk_addr += sit_i->sit_blocks; | |
685 | ||
686 | return blk_addr; | |
687 | } | |
688 | ||
689 | static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, | |
690 | pgoff_t block_addr) | |
691 | { | |
692 | struct sit_info *sit_i = SIT_I(sbi); | |
693 | block_addr -= sit_i->sit_base_addr; | |
694 | if (block_addr < sit_i->sit_blocks) | |
695 | block_addr += sit_i->sit_blocks; | |
696 | else | |
697 | block_addr -= sit_i->sit_blocks; | |
698 | ||
699 | return block_addr + sit_i->sit_base_addr; | |
700 | } | |
701 | ||
702 | static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) | |
703 | { | |
d3a14afd | 704 | unsigned int block_off = SIT_BLOCK_OFFSET(start); |
39a53e0c | 705 | |
c6ac4c0e | 706 | f2fs_change_bit(block_off, sit_i->sit_bitmap); |
ae27d62e CY |
707 | #ifdef CONFIG_F2FS_CHECK_FS |
708 | f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); | |
709 | #endif | |
39a53e0c JK |
710 | } |
711 | ||
712 | static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi) | |
713 | { | |
714 | struct sit_info *sit_i = SIT_I(sbi); | |
48fbfe50 DD |
715 | time64_t now = ktime_get_real_seconds(); |
716 | ||
717 | return sit_i->elapsed_time + now - sit_i->mounted_time; | |
39a53e0c JK |
718 | } |
719 | ||
720 | static inline void set_summary(struct f2fs_summary *sum, nid_t nid, | |
721 | unsigned int ofs_in_node, unsigned char version) | |
722 | { | |
723 | sum->nid = cpu_to_le32(nid); | |
724 | sum->ofs_in_node = cpu_to_le16(ofs_in_node); | |
725 | sum->version = version; | |
726 | } | |
727 | ||
728 | static inline block_t start_sum_block(struct f2fs_sb_info *sbi) | |
729 | { | |
730 | return __start_cp_addr(sbi) + | |
731 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); | |
732 | } | |
733 | ||
734 | static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) | |
735 | { | |
736 | return __start_cp_addr(sbi) + | |
737 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) | |
738 | - (base + 1) + type; | |
739 | } | |
5ec4e49f | 740 | |
e93b9865 HP |
741 | static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi, |
742 | unsigned int secno) | |
743 | { | |
4ddb1a4d | 744 | if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >= |
302bd348 | 745 | sbi->fggc_threshold) |
e93b9865 HP |
746 | return true; |
747 | return false; | |
748 | } | |
749 | ||
5ec4e49f JK |
750 | static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) |
751 | { | |
752 | if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) | |
753 | return true; | |
754 | return false; | |
755 | } | |
ac5d156c | 756 | |
87d6f890 JK |
757 | /* |
758 | * It is very important to gather dirty pages and write at once, so that we can | |
759 | * submit a big bio without interfering other data writes. | |
760 | * By default, 512 pages for directory data, | |
727ebb09 KM |
761 | * 512 pages (2MB) * 8 for nodes, and |
762 | * 256 pages * 8 for meta are set. | |
87d6f890 JK |
763 | */ |
764 | static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) | |
765 | { | |
a88a341a | 766 | if (sbi->sb->s_bdi->wb.dirty_exceeded) |
510184c8 JK |
767 | return 0; |
768 | ||
a1257023 JK |
769 | if (type == DATA) |
770 | return sbi->blocks_per_seg; | |
771 | else if (type == NODE) | |
2c237eba | 772 | return 8 * sbi->blocks_per_seg; |
87d6f890 | 773 | else if (type == META) |
664ba972 | 774 | return 8 * BIO_MAX_PAGES; |
87d6f890 JK |
775 | else |
776 | return 0; | |
777 | } | |
50c8cdb3 JK |
778 | |
779 | /* | |
780 | * When writing pages, it'd better align nr_to_write for segment size. | |
781 | */ | |
782 | static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, | |
783 | struct writeback_control *wbc) | |
784 | { | |
785 | long nr_to_write, desired; | |
786 | ||
787 | if (wbc->sync_mode != WB_SYNC_NONE) | |
788 | return 0; | |
789 | ||
790 | nr_to_write = wbc->nr_to_write; | |
664ba972 | 791 | desired = BIO_MAX_PAGES; |
28ea6162 | 792 | if (type == NODE) |
664ba972 | 793 | desired <<= 1; |
50c8cdb3 JK |
794 | |
795 | wbc->nr_to_write = desired; | |
796 | return desired - nr_to_write; | |
797 | } |