]>
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, | |
141 | GC_GREEDY | |
142 | }; | |
143 | ||
144 | /* | |
145 | * BG_GC means the background cleaning job. | |
146 | * FG_GC means the on-demand cleaning job. | |
6aefd93b | 147 | * FORCE_FG_GC means on-demand cleaning job in background. |
39a53e0c JK |
148 | */ |
149 | enum { | |
150 | BG_GC = 0, | |
6aefd93b JK |
151 | FG_GC, |
152 | FORCE_FG_GC, | |
39a53e0c JK |
153 | }; |
154 | ||
155 | /* for a function parameter to select a victim segment */ | |
156 | struct victim_sel_policy { | |
157 | int alloc_mode; /* LFS or SSR */ | |
158 | int gc_mode; /* GC_CB or GC_GREEDY */ | |
159 | unsigned long *dirty_segmap; /* dirty segment bitmap */ | |
a26b7c8a | 160 | unsigned int max_search; /* maximum # of segments to search */ |
39a53e0c JK |
161 | unsigned int offset; /* last scanned bitmap offset */ |
162 | unsigned int ofs_unit; /* bitmap search unit */ | |
163 | unsigned int min_cost; /* minimum cost */ | |
164 | unsigned int min_segno; /* segment # having min. cost */ | |
165 | }; | |
166 | ||
167 | struct seg_entry { | |
f51b4ce6 CY |
168 | unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ |
169 | unsigned int valid_blocks:10; /* # of valid blocks */ | |
170 | unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ | |
171 | unsigned int padding:6; /* padding */ | |
39a53e0c | 172 | unsigned char *cur_valid_map; /* validity bitmap of blocks */ |
355e7891 CY |
173 | #ifdef CONFIG_F2FS_CHECK_FS |
174 | unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ | |
175 | #endif | |
39a53e0c JK |
176 | /* |
177 | * # of valid blocks and the validity bitmap stored in the the last | |
178 | * checkpoint pack. This information is used by the SSR mode. | |
179 | */ | |
f51b4ce6 | 180 | unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ |
a66cdd98 | 181 | unsigned char *discard_map; |
39a53e0c JK |
182 | unsigned long long mtime; /* modification time of the segment */ |
183 | }; | |
184 | ||
185 | struct sec_entry { | |
186 | unsigned int valid_blocks; /* # of valid blocks in a section */ | |
187 | }; | |
188 | ||
189 | struct segment_allocation { | |
190 | void (*allocate_segment)(struct f2fs_sb_info *, int, bool); | |
191 | }; | |
192 | ||
decd36b6 CY |
193 | /* |
194 | * this value is set in page as a private data which indicate that | |
195 | * the page is atomically written, and it is in inmem_pages list. | |
196 | */ | |
d48dfc20 | 197 | #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1) |
0a595eba | 198 | #define DUMMY_WRITTEN_PAGE ((unsigned long)-2) |
decd36b6 CY |
199 | |
200 | #define IS_ATOMIC_WRITTEN_PAGE(page) \ | |
201 | (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE) | |
0a595eba JK |
202 | #define IS_DUMMY_WRITTEN_PAGE(page) \ |
203 | (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE) | |
decd36b6 | 204 | |
88b88a66 JK |
205 | struct inmem_pages { |
206 | struct list_head list; | |
207 | struct page *page; | |
28bc106b | 208 | block_t old_addr; /* for revoking when fail to commit */ |
88b88a66 JK |
209 | }; |
210 | ||
39a53e0c JK |
211 | struct sit_info { |
212 | const struct segment_allocation *s_ops; | |
213 | ||
214 | block_t sit_base_addr; /* start block address of SIT area */ | |
215 | block_t sit_blocks; /* # of blocks used by SIT area */ | |
216 | block_t written_valid_blocks; /* # of valid blocks in main area */ | |
217 | char *sit_bitmap; /* SIT bitmap pointer */ | |
ae27d62e CY |
218 | #ifdef CONFIG_F2FS_CHECK_FS |
219 | char *sit_bitmap_mir; /* SIT bitmap mirror */ | |
220 | #endif | |
39a53e0c JK |
221 | unsigned int bitmap_size; /* SIT bitmap size */ |
222 | ||
60a3b782 | 223 | unsigned long *tmp_map; /* bitmap for temporal use */ |
39a53e0c JK |
224 | unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ |
225 | unsigned int dirty_sentries; /* # of dirty sentries */ | |
226 | unsigned int sents_per_block; /* # of SIT entries per block */ | |
227 | struct mutex sentry_lock; /* to protect SIT cache */ | |
228 | struct seg_entry *sentries; /* SIT segment-level cache */ | |
229 | struct sec_entry *sec_entries; /* SIT section-level cache */ | |
230 | ||
231 | /* for cost-benefit algorithm in cleaning procedure */ | |
232 | unsigned long long elapsed_time; /* elapsed time after mount */ | |
233 | unsigned long long mounted_time; /* mount time */ | |
234 | unsigned long long min_mtime; /* min. modification time */ | |
235 | unsigned long long max_mtime; /* max. modification time */ | |
236 | }; | |
237 | ||
238 | struct free_segmap_info { | |
239 | unsigned int start_segno; /* start segment number logically */ | |
240 | unsigned int free_segments; /* # of free segments */ | |
241 | unsigned int free_sections; /* # of free sections */ | |
1a118ccf | 242 | spinlock_t segmap_lock; /* free segmap lock */ |
39a53e0c JK |
243 | unsigned long *free_segmap; /* free segment bitmap */ |
244 | unsigned long *free_secmap; /* free section bitmap */ | |
245 | }; | |
246 | ||
247 | /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ | |
248 | enum dirty_type { | |
249 | DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ | |
250 | DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ | |
251 | DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ | |
252 | DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ | |
253 | DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ | |
254 | DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ | |
255 | DIRTY, /* to count # of dirty segments */ | |
256 | PRE, /* to count # of entirely obsolete segments */ | |
257 | NR_DIRTY_TYPE | |
258 | }; | |
259 | ||
260 | struct dirty_seglist_info { | |
261 | const struct victim_selection *v_ops; /* victim selction operation */ | |
262 | unsigned long *dirty_segmap[NR_DIRTY_TYPE]; | |
263 | struct mutex seglist_lock; /* lock for segment bitmaps */ | |
264 | int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ | |
5ec4e49f | 265 | unsigned long *victim_secmap; /* background GC victims */ |
39a53e0c JK |
266 | }; |
267 | ||
268 | /* victim selection function for cleaning and SSR */ | |
269 | struct victim_selection { | |
270 | int (*get_victim)(struct f2fs_sb_info *, unsigned int *, | |
271 | int, int, char); | |
272 | }; | |
273 | ||
274 | /* for active log information */ | |
275 | struct curseg_info { | |
276 | struct mutex curseg_mutex; /* lock for consistency */ | |
277 | struct f2fs_summary_block *sum_blk; /* cached summary block */ | |
b7ad7512 CY |
278 | struct rw_semaphore journal_rwsem; /* protect journal area */ |
279 | struct f2fs_journal *journal; /* cached journal info */ | |
39a53e0c JK |
280 | unsigned char alloc_type; /* current allocation type */ |
281 | unsigned int segno; /* current segment number */ | |
282 | unsigned short next_blkoff; /* next block offset to write */ | |
283 | unsigned int zone; /* current zone number */ | |
284 | unsigned int next_segno; /* preallocated segment */ | |
285 | }; | |
286 | ||
184a5cd2 CY |
287 | struct sit_entry_set { |
288 | struct list_head set_list; /* link with all sit sets */ | |
289 | unsigned int start_segno; /* start segno of sits in set */ | |
290 | unsigned int entry_cnt; /* the # of sit entries in set */ | |
291 | }; | |
292 | ||
39a53e0c JK |
293 | /* |
294 | * inline functions | |
295 | */ | |
296 | static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) | |
297 | { | |
298 | return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); | |
299 | } | |
300 | ||
301 | static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, | |
302 | unsigned int segno) | |
303 | { | |
304 | struct sit_info *sit_i = SIT_I(sbi); | |
305 | return &sit_i->sentries[segno]; | |
306 | } | |
307 | ||
308 | static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, | |
309 | unsigned int segno) | |
310 | { | |
311 | struct sit_info *sit_i = SIT_I(sbi); | |
4ddb1a4d | 312 | return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; |
39a53e0c JK |
313 | } |
314 | ||
315 | static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, | |
302bd348 | 316 | unsigned int segno, bool use_section) |
39a53e0c JK |
317 | { |
318 | /* | |
319 | * In order to get # of valid blocks in a section instantly from many | |
320 | * segments, f2fs manages two counting structures separately. | |
321 | */ | |
302bd348 | 322 | if (use_section && sbi->segs_per_sec > 1) |
39a53e0c JK |
323 | return get_sec_entry(sbi, segno)->valid_blocks; |
324 | else | |
325 | return get_seg_entry(sbi, segno)->valid_blocks; | |
326 | } | |
327 | ||
328 | static inline void seg_info_from_raw_sit(struct seg_entry *se, | |
329 | struct f2fs_sit_entry *rs) | |
330 | { | |
331 | se->valid_blocks = GET_SIT_VBLOCKS(rs); | |
332 | se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); | |
333 | memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
334 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
355e7891 CY |
335 | #ifdef CONFIG_F2FS_CHECK_FS |
336 | memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
337 | #endif | |
39a53e0c JK |
338 | se->type = GET_SIT_TYPE(rs); |
339 | se->mtime = le64_to_cpu(rs->mtime); | |
340 | } | |
341 | ||
342 | static inline void seg_info_to_raw_sit(struct seg_entry *se, | |
343 | struct f2fs_sit_entry *rs) | |
344 | { | |
345 | unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | | |
346 | se->valid_blocks; | |
347 | rs->vblocks = cpu_to_le16(raw_vblocks); | |
348 | memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); | |
349 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
350 | se->ckpt_valid_blocks = se->valid_blocks; | |
351 | rs->mtime = cpu_to_le64(se->mtime); | |
352 | } | |
353 | ||
354 | static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, | |
355 | unsigned int max, unsigned int segno) | |
356 | { | |
357 | unsigned int ret; | |
1a118ccf | 358 | spin_lock(&free_i->segmap_lock); |
39a53e0c | 359 | ret = find_next_bit(free_i->free_segmap, max, segno); |
1a118ccf | 360 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
361 | return ret; |
362 | } | |
363 | ||
364 | static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) | |
365 | { | |
366 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
367 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
368 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
369 | unsigned int next; |
370 | ||
1a118ccf | 371 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
372 | clear_bit(segno, free_i->free_segmap); |
373 | free_i->free_segments++; | |
374 | ||
7fd97019 WL |
375 | next = find_next_bit(free_i->free_segmap, |
376 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
377 | if (next >= start_segno + sbi->segs_per_sec) { |
378 | clear_bit(secno, free_i->free_secmap); | |
379 | free_i->free_sections++; | |
380 | } | |
1a118ccf | 381 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
382 | } |
383 | ||
384 | static inline void __set_inuse(struct f2fs_sb_info *sbi, | |
385 | unsigned int segno) | |
386 | { | |
387 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
388 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
389 | ||
39a53e0c JK |
390 | set_bit(segno, free_i->free_segmap); |
391 | free_i->free_segments--; | |
392 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
393 | free_i->free_sections--; | |
394 | } | |
395 | ||
396 | static inline void __set_test_and_free(struct f2fs_sb_info *sbi, | |
397 | unsigned int segno) | |
398 | { | |
399 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
400 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
401 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
402 | unsigned int next; |
403 | ||
1a118ccf | 404 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
405 | if (test_and_clear_bit(segno, free_i->free_segmap)) { |
406 | free_i->free_segments++; | |
407 | ||
f1121ab0 CY |
408 | next = find_next_bit(free_i->free_segmap, |
409 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
410 | if (next >= start_segno + sbi->segs_per_sec) { |
411 | if (test_and_clear_bit(secno, free_i->free_secmap)) | |
412 | free_i->free_sections++; | |
413 | } | |
414 | } | |
1a118ccf | 415 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
416 | } |
417 | ||
418 | static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, | |
419 | unsigned int segno) | |
420 | { | |
421 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
422 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
423 | ||
1a118ccf | 424 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
425 | if (!test_and_set_bit(segno, free_i->free_segmap)) { |
426 | free_i->free_segments--; | |
427 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
428 | free_i->free_sections--; | |
429 | } | |
1a118ccf | 430 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
431 | } |
432 | ||
433 | static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, | |
434 | void *dst_addr) | |
435 | { | |
436 | struct sit_info *sit_i = SIT_I(sbi); | |
ae27d62e CY |
437 | |
438 | #ifdef CONFIG_F2FS_CHECK_FS | |
439 | if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, | |
440 | sit_i->bitmap_size)) | |
441 | f2fs_bug_on(sbi, 1); | |
442 | #endif | |
39a53e0c JK |
443 | memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); |
444 | } | |
445 | ||
446 | static inline block_t written_block_count(struct f2fs_sb_info *sbi) | |
447 | { | |
8b8343fa | 448 | return SIT_I(sbi)->written_valid_blocks; |
39a53e0c JK |
449 | } |
450 | ||
451 | static inline unsigned int free_segments(struct f2fs_sb_info *sbi) | |
452 | { | |
8b8343fa | 453 | return FREE_I(sbi)->free_segments; |
39a53e0c JK |
454 | } |
455 | ||
456 | static inline int reserved_segments(struct f2fs_sb_info *sbi) | |
457 | { | |
458 | return SM_I(sbi)->reserved_segments; | |
459 | } | |
460 | ||
461 | static inline unsigned int free_sections(struct f2fs_sb_info *sbi) | |
462 | { | |
8b8343fa | 463 | return FREE_I(sbi)->free_sections; |
39a53e0c JK |
464 | } |
465 | ||
466 | static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) | |
467 | { | |
468 | return DIRTY_I(sbi)->nr_dirty[PRE]; | |
469 | } | |
470 | ||
471 | static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) | |
472 | { | |
473 | return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + | |
474 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + | |
475 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + | |
476 | DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + | |
477 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + | |
478 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; | |
479 | } | |
480 | ||
481 | static inline int overprovision_segments(struct f2fs_sb_info *sbi) | |
482 | { | |
483 | return SM_I(sbi)->ovp_segments; | |
484 | } | |
485 | ||
486 | static inline int overprovision_sections(struct f2fs_sb_info *sbi) | |
487 | { | |
4ddb1a4d | 488 | return GET_SEC_FROM_SEG(sbi, (unsigned int)overprovision_segments(sbi)); |
39a53e0c JK |
489 | } |
490 | ||
491 | static inline int reserved_sections(struct f2fs_sb_info *sbi) | |
492 | { | |
4ddb1a4d | 493 | return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi)); |
39a53e0c JK |
494 | } |
495 | ||
496 | static inline bool need_SSR(struct f2fs_sb_info *sbi) | |
497 | { | |
95dd8973 JK |
498 | int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); |
499 | int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); | |
b9610bdf | 500 | int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA); |
36abef4e JK |
501 | |
502 | if (test_opt(sbi, LFS)) | |
503 | return false; | |
504 | ||
b9610bdf | 505 | return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs + |
796dbbfe | 506 | 2 * reserved_sections(sbi)); |
39a53e0c JK |
507 | } |
508 | ||
7f3037a5 JK |
509 | static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, |
510 | int freed, int needed) | |
39a53e0c | 511 | { |
5ac206cf NJ |
512 | int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); |
513 | int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); | |
b9610bdf | 514 | int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA); |
0f18b462 | 515 | |
caf0047e | 516 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
029cd28c JK |
517 | return false; |
518 | ||
7f3037a5 | 519 | return (free_sections(sbi) + freed) <= |
b9610bdf JK |
520 | (node_secs + 2 * dent_secs + imeta_secs + |
521 | reserved_sections(sbi) + needed); | |
39a53e0c JK |
522 | } |
523 | ||
81eb8d6e JK |
524 | static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) |
525 | { | |
6c311ec6 | 526 | return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; |
81eb8d6e JK |
527 | } |
528 | ||
39a53e0c JK |
529 | static inline int utilization(struct f2fs_sb_info *sbi) |
530 | { | |
6c311ec6 CF |
531 | return div_u64((u64)valid_user_blocks(sbi) * 100, |
532 | sbi->user_block_count); | |
39a53e0c JK |
533 | } |
534 | ||
535 | /* | |
536 | * Sometimes f2fs may be better to drop out-of-place update policy. | |
216fbd64 JK |
537 | * And, users can control the policy through sysfs entries. |
538 | * There are five policies with triggering conditions as follows. | |
539 | * F2FS_IPU_FORCE - all the time, | |
540 | * F2FS_IPU_SSR - if SSR mode is activated, | |
541 | * F2FS_IPU_UTIL - if FS utilization is over threashold, | |
542 | * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over | |
543 | * threashold, | |
c1ce1b02 JK |
544 | * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash |
545 | * storages. IPU will be triggered only if the # of dirty | |
546 | * pages over min_fsync_blocks. | |
216fbd64 | 547 | * F2FS_IPUT_DISABLE - disable IPU. (=default option) |
39a53e0c | 548 | */ |
216fbd64 | 549 | #define DEF_MIN_IPU_UTIL 70 |
c1ce1b02 | 550 | #define DEF_MIN_FSYNC_BLOCKS 8 |
ef095d19 | 551 | #define DEF_MIN_HOT_BLOCKS 16 |
216fbd64 JK |
552 | |
553 | enum { | |
554 | F2FS_IPU_FORCE, | |
555 | F2FS_IPU_SSR, | |
556 | F2FS_IPU_UTIL, | |
557 | F2FS_IPU_SSR_UTIL, | |
c1ce1b02 | 558 | F2FS_IPU_FSYNC, |
04485987 | 559 | F2FS_IPU_ASYNC, |
216fbd64 JK |
560 | }; |
561 | ||
04485987 HP |
562 | static inline bool need_inplace_update(struct inode *inode, |
563 | struct f2fs_io_info *fio) | |
39a53e0c | 564 | { |
4081363f | 565 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9b5f136f | 566 | unsigned int policy = SM_I(sbi)->ipu_policy; |
216fbd64 JK |
567 | |
568 | /* IPU can be done only for the user data */ | |
88b88a66 | 569 | if (S_ISDIR(inode->i_mode) || f2fs_is_atomic_file(inode)) |
39a53e0c | 570 | return false; |
216fbd64 | 571 | |
36abef4e JK |
572 | if (test_opt(sbi, LFS)) |
573 | return false; | |
574 | ||
9b5f136f | 575 | if (policy & (0x1 << F2FS_IPU_FORCE)) |
39a53e0c | 576 | return true; |
9b5f136f JK |
577 | if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi)) |
578 | return true; | |
579 | if (policy & (0x1 << F2FS_IPU_UTIL) && | |
580 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
581 | return true; | |
582 | if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) && | |
583 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
584 | return true; | |
585 | ||
04485987 HP |
586 | /* |
587 | * IPU for rewrite async pages | |
588 | */ | |
589 | if (policy & (0x1 << F2FS_IPU_ASYNC) && | |
590 | fio && fio->op == REQ_OP_WRITE && | |
591 | !(fio->op_flags & REQ_SYNC)) | |
592 | return true; | |
593 | ||
9b5f136f JK |
594 | /* this is only set during fdatasync */ |
595 | if (policy & (0x1 << F2FS_IPU_FSYNC) && | |
91942321 | 596 | is_inode_flag_set(inode, FI_NEED_IPU)) |
9b5f136f JK |
597 | return true; |
598 | ||
39a53e0c JK |
599 | return false; |
600 | } | |
601 | ||
602 | static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, | |
603 | int type) | |
604 | { | |
605 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
606 | return curseg->segno; | |
607 | } | |
608 | ||
609 | static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, | |
610 | int type) | |
611 | { | |
612 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
613 | return curseg->alloc_type; | |
614 | } | |
615 | ||
616 | static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type) | |
617 | { | |
618 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
619 | return curseg->next_blkoff; | |
620 | } | |
621 | ||
622 | static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) | |
623 | { | |
7a04f64d | 624 | f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1); |
39a53e0c JK |
625 | } |
626 | ||
39a53e0c JK |
627 | static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr) |
628 | { | |
bb413d6a YH |
629 | BUG_ON(blk_addr < SEG0_BLKADDR(sbi) |
630 | || blk_addr >= MAX_BLKADDR(sbi)); | |
39a53e0c JK |
631 | } |
632 | ||
633 | /* | |
e1c42045 | 634 | * Summary block is always treated as an invalid block |
39a53e0c JK |
635 | */ |
636 | static inline void check_block_count(struct f2fs_sb_info *sbi, | |
637 | int segno, struct f2fs_sit_entry *raw_sit) | |
638 | { | |
4c278394 | 639 | #ifdef CONFIG_F2FS_CHECK_FS |
44c60bf2 | 640 | bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; |
39a53e0c | 641 | int valid_blocks = 0; |
44c60bf2 | 642 | int cur_pos = 0, next_pos; |
39a53e0c | 643 | |
39a53e0c | 644 | /* check bitmap with valid block count */ |
44c60bf2 CY |
645 | do { |
646 | if (is_valid) { | |
647 | next_pos = find_next_zero_bit_le(&raw_sit->valid_map, | |
648 | sbi->blocks_per_seg, | |
649 | cur_pos); | |
650 | valid_blocks += next_pos - cur_pos; | |
651 | } else | |
652 | next_pos = find_next_bit_le(&raw_sit->valid_map, | |
653 | sbi->blocks_per_seg, | |
654 | cur_pos); | |
655 | cur_pos = next_pos; | |
656 | is_valid = !is_valid; | |
657 | } while (cur_pos < sbi->blocks_per_seg); | |
39a53e0c | 658 | BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks); |
5d56b671 | 659 | #endif |
4c278394 JK |
660 | /* check segment usage, and check boundary of a given segment number */ |
661 | f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg | |
662 | || segno > TOTAL_SEGS(sbi) - 1); | |
7a04f64d | 663 | } |
39a53e0c JK |
664 | |
665 | static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, | |
666 | unsigned int start) | |
667 | { | |
668 | struct sit_info *sit_i = SIT_I(sbi); | |
d3a14afd | 669 | unsigned int offset = SIT_BLOCK_OFFSET(start); |
39a53e0c JK |
670 | block_t blk_addr = sit_i->sit_base_addr + offset; |
671 | ||
672 | check_seg_range(sbi, start); | |
673 | ||
ae27d62e CY |
674 | #ifdef CONFIG_F2FS_CHECK_FS |
675 | if (f2fs_test_bit(offset, sit_i->sit_bitmap) != | |
676 | f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) | |
677 | f2fs_bug_on(sbi, 1); | |
678 | #endif | |
679 | ||
39a53e0c JK |
680 | /* calculate sit block address */ |
681 | if (f2fs_test_bit(offset, sit_i->sit_bitmap)) | |
682 | blk_addr += sit_i->sit_blocks; | |
683 | ||
684 | return blk_addr; | |
685 | } | |
686 | ||
687 | static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, | |
688 | pgoff_t block_addr) | |
689 | { | |
690 | struct sit_info *sit_i = SIT_I(sbi); | |
691 | block_addr -= sit_i->sit_base_addr; | |
692 | if (block_addr < sit_i->sit_blocks) | |
693 | block_addr += sit_i->sit_blocks; | |
694 | else | |
695 | block_addr -= sit_i->sit_blocks; | |
696 | ||
697 | return block_addr + sit_i->sit_base_addr; | |
698 | } | |
699 | ||
700 | static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) | |
701 | { | |
d3a14afd | 702 | unsigned int block_off = SIT_BLOCK_OFFSET(start); |
39a53e0c | 703 | |
c6ac4c0e | 704 | f2fs_change_bit(block_off, sit_i->sit_bitmap); |
ae27d62e CY |
705 | #ifdef CONFIG_F2FS_CHECK_FS |
706 | f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); | |
707 | #endif | |
39a53e0c JK |
708 | } |
709 | ||
710 | static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi) | |
711 | { | |
712 | struct sit_info *sit_i = SIT_I(sbi); | |
713 | return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec - | |
714 | sit_i->mounted_time; | |
715 | } | |
716 | ||
717 | static inline void set_summary(struct f2fs_summary *sum, nid_t nid, | |
718 | unsigned int ofs_in_node, unsigned char version) | |
719 | { | |
720 | sum->nid = cpu_to_le32(nid); | |
721 | sum->ofs_in_node = cpu_to_le16(ofs_in_node); | |
722 | sum->version = version; | |
723 | } | |
724 | ||
725 | static inline block_t start_sum_block(struct f2fs_sb_info *sbi) | |
726 | { | |
727 | return __start_cp_addr(sbi) + | |
728 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); | |
729 | } | |
730 | ||
731 | static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) | |
732 | { | |
733 | return __start_cp_addr(sbi) + | |
734 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) | |
735 | - (base + 1) + type; | |
736 | } | |
5ec4e49f | 737 | |
e93b9865 HP |
738 | static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi, |
739 | unsigned int secno) | |
740 | { | |
4ddb1a4d | 741 | if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >= |
302bd348 | 742 | sbi->fggc_threshold) |
e93b9865 HP |
743 | return true; |
744 | return false; | |
745 | } | |
746 | ||
5ec4e49f JK |
747 | static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) |
748 | { | |
749 | if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) | |
750 | return true; | |
751 | return false; | |
752 | } | |
ac5d156c | 753 | |
87d6f890 JK |
754 | /* |
755 | * It is very important to gather dirty pages and write at once, so that we can | |
756 | * submit a big bio without interfering other data writes. | |
757 | * By default, 512 pages for directory data, | |
727ebb09 KM |
758 | * 512 pages (2MB) * 8 for nodes, and |
759 | * 256 pages * 8 for meta are set. | |
87d6f890 JK |
760 | */ |
761 | static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) | |
762 | { | |
a88a341a | 763 | if (sbi->sb->s_bdi->wb.dirty_exceeded) |
510184c8 JK |
764 | return 0; |
765 | ||
a1257023 JK |
766 | if (type == DATA) |
767 | return sbi->blocks_per_seg; | |
768 | else if (type == NODE) | |
2c237eba | 769 | return 8 * sbi->blocks_per_seg; |
87d6f890 | 770 | else if (type == META) |
664ba972 | 771 | return 8 * BIO_MAX_PAGES; |
87d6f890 JK |
772 | else |
773 | return 0; | |
774 | } | |
50c8cdb3 JK |
775 | |
776 | /* | |
777 | * When writing pages, it'd better align nr_to_write for segment size. | |
778 | */ | |
779 | static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, | |
780 | struct writeback_control *wbc) | |
781 | { | |
782 | long nr_to_write, desired; | |
783 | ||
784 | if (wbc->sync_mode != WB_SYNC_NONE) | |
785 | return 0; | |
786 | ||
787 | nr_to_write = wbc->nr_to_write; | |
664ba972 | 788 | desired = BIO_MAX_PAGES; |
28ea6162 | 789 | if (type == NODE) |
664ba972 | 790 | desired <<= 1; |
50c8cdb3 JK |
791 | |
792 | wbc->nr_to_write = desired; | |
793 | return desired - nr_to_write; | |
794 | } |