]>
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 | |
a912b54d JK |
30 | #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA) |
31 | #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA) | |
32 | #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA) | |
33 | ||
5c773ba3 | 34 | #define IS_CURSEG(sbi, seg) \ |
68afcf2d TK |
35 | (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \ |
36 | ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \ | |
37 | ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \ | |
38 | ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \ | |
39 | ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \ | |
40 | ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno)) | |
39a53e0c JK |
41 | |
42 | #define IS_CURSEC(sbi, secno) \ | |
68afcf2d TK |
43 | (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \ |
44 | (sbi)->segs_per_sec) || \ | |
45 | ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \ | |
46 | (sbi)->segs_per_sec) || \ | |
47 | ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \ | |
48 | (sbi)->segs_per_sec) || \ | |
49 | ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \ | |
50 | (sbi)->segs_per_sec) || \ | |
51 | ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \ | |
52 | (sbi)->segs_per_sec) || \ | |
53 | ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \ | |
54 | (sbi)->segs_per_sec)) \ | |
39a53e0c | 55 | |
7cd8558b JK |
56 | #define MAIN_BLKADDR(sbi) (SM_I(sbi)->main_blkaddr) |
57 | #define SEG0_BLKADDR(sbi) (SM_I(sbi)->seg0_blkaddr) | |
58 | ||
59 | #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments) | |
68afcf2d | 60 | #define MAIN_SECS(sbi) ((sbi)->total_sections) |
7cd8558b JK |
61 | |
62 | #define TOTAL_SEGS(sbi) (SM_I(sbi)->segment_count) | |
68afcf2d | 63 | #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg) |
7cd8558b JK |
64 | |
65 | #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi)) | |
68afcf2d TK |
66 | #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \ |
67 | (sbi)->log_blocks_per_seg)) | |
7cd8558b JK |
68 | |
69 | #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \ | |
68afcf2d | 70 | (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg)) |
7cd8558b | 71 | |
39a53e0c | 72 | #define NEXT_FREE_BLKADDR(sbi, curseg) \ |
68afcf2d | 73 | (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff) |
39a53e0c | 74 | |
7cd8558b | 75 | #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi)) |
39a53e0c | 76 | #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ |
68afcf2d | 77 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg) |
491c0854 | 78 | #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ |
68afcf2d | 79 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1)) |
491c0854 | 80 | |
39a53e0c | 81 | #define GET_SEGNO(sbi, blk_addr) \ |
68afcf2d | 82 | ((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ? \ |
39a53e0c JK |
83 | NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ |
84 | GET_SEGNO_FROM_SEG0(sbi, blk_addr))) | |
4ddb1a4d JK |
85 | #define BLKS_PER_SEC(sbi) \ |
86 | ((sbi)->segs_per_sec * (sbi)->blocks_per_seg) | |
87 | #define GET_SEC_FROM_SEG(sbi, segno) \ | |
68afcf2d | 88 | ((segno) / (sbi)->segs_per_sec) |
4ddb1a4d | 89 | #define GET_SEG_FROM_SEC(sbi, secno) \ |
63fcf8e8 | 90 | ((secno) * (sbi)->segs_per_sec) |
4ddb1a4d JK |
91 | #define GET_ZONE_FROM_SEC(sbi, secno) \ |
92 | ((secno) / (sbi)->secs_per_zone) | |
93 | #define GET_ZONE_FROM_SEG(sbi, segno) \ | |
94 | GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) | |
39a53e0c JK |
95 | |
96 | #define GET_SUM_BLOCK(sbi, segno) \ | |
68afcf2d | 97 | ((sbi)->sm_info->ssa_blkaddr + (segno)) |
39a53e0c JK |
98 | |
99 | #define GET_SUM_TYPE(footer) ((footer)->entry_type) | |
68afcf2d | 100 | #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type)) |
39a53e0c JK |
101 | |
102 | #define SIT_ENTRY_OFFSET(sit_i, segno) \ | |
68afcf2d | 103 | ((segno) % (sit_i)->sents_per_block) |
d3a14afd | 104 | #define SIT_BLOCK_OFFSET(segno) \ |
68afcf2d | 105 | ((segno) / SIT_ENTRY_PER_BLOCK) |
d3a14afd CY |
106 | #define START_SEGNO(segno) \ |
107 | (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK) | |
74de593a | 108 | #define SIT_BLK_CNT(sbi) \ |
7cd8558b | 109 | ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK) |
39a53e0c JK |
110 | #define f2fs_bitmap_size(nr) \ |
111 | (BITS_TO_LONGS(nr) * sizeof(unsigned long)) | |
39a53e0c | 112 | |
55cf9cb6 CY |
113 | #define SECTOR_FROM_BLOCK(blk_addr) \ |
114 | (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK) | |
115 | #define SECTOR_TO_BLOCK(sectors) \ | |
68afcf2d | 116 | ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK) |
3cd8a239 | 117 | |
39a53e0c JK |
118 | /* |
119 | * indicate a block allocation direction: RIGHT and LEFT. | |
120 | * RIGHT means allocating new sections towards the end of volume. | |
121 | * LEFT means the opposite direction. | |
122 | */ | |
123 | enum { | |
124 | ALLOC_RIGHT = 0, | |
125 | ALLOC_LEFT | |
126 | }; | |
127 | ||
128 | /* | |
129 | * In the victim_sel_policy->alloc_mode, there are two block allocation modes. | |
130 | * LFS writes data sequentially with cleaning operations. | |
131 | * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. | |
132 | */ | |
133 | enum { | |
134 | LFS = 0, | |
135 | SSR | |
136 | }; | |
137 | ||
138 | /* | |
139 | * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes. | |
140 | * GC_CB is based on cost-benefit algorithm. | |
141 | * GC_GREEDY is based on greedy algorithm. | |
142 | */ | |
143 | enum { | |
144 | GC_CB = 0, | |
e066b83c JK |
145 | GC_GREEDY, |
146 | ALLOC_NEXT, | |
147 | FLUSH_DEVICE, | |
148 | MAX_GC_POLICY, | |
39a53e0c JK |
149 | }; |
150 | ||
151 | /* | |
152 | * BG_GC means the background cleaning job. | |
153 | * FG_GC means the on-demand cleaning job. | |
6aefd93b | 154 | * FORCE_FG_GC means on-demand cleaning job in background. |
39a53e0c JK |
155 | */ |
156 | enum { | |
157 | BG_GC = 0, | |
6aefd93b JK |
158 | FG_GC, |
159 | FORCE_FG_GC, | |
39a53e0c JK |
160 | }; |
161 | ||
162 | /* for a function parameter to select a victim segment */ | |
163 | struct victim_sel_policy { | |
164 | int alloc_mode; /* LFS or SSR */ | |
165 | int gc_mode; /* GC_CB or GC_GREEDY */ | |
166 | unsigned long *dirty_segmap; /* dirty segment bitmap */ | |
a26b7c8a | 167 | unsigned int max_search; /* maximum # of segments to search */ |
39a53e0c JK |
168 | unsigned int offset; /* last scanned bitmap offset */ |
169 | unsigned int ofs_unit; /* bitmap search unit */ | |
170 | unsigned int min_cost; /* minimum cost */ | |
171 | unsigned int min_segno; /* segment # having min. cost */ | |
172 | }; | |
173 | ||
174 | struct seg_entry { | |
f51b4ce6 CY |
175 | unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ |
176 | unsigned int valid_blocks:10; /* # of valid blocks */ | |
177 | unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ | |
178 | unsigned int padding:6; /* padding */ | |
39a53e0c | 179 | unsigned char *cur_valid_map; /* validity bitmap of blocks */ |
355e7891 CY |
180 | #ifdef CONFIG_F2FS_CHECK_FS |
181 | unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ | |
182 | #endif | |
39a53e0c JK |
183 | /* |
184 | * # of valid blocks and the validity bitmap stored in the the last | |
185 | * checkpoint pack. This information is used by the SSR mode. | |
186 | */ | |
f51b4ce6 | 187 | unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ |
a66cdd98 | 188 | unsigned char *discard_map; |
39a53e0c JK |
189 | unsigned long long mtime; /* modification time of the segment */ |
190 | }; | |
191 | ||
192 | struct sec_entry { | |
193 | unsigned int valid_blocks; /* # of valid blocks in a section */ | |
194 | }; | |
195 | ||
196 | struct segment_allocation { | |
197 | void (*allocate_segment)(struct f2fs_sb_info *, int, bool); | |
198 | }; | |
199 | ||
decd36b6 CY |
200 | /* |
201 | * this value is set in page as a private data which indicate that | |
202 | * the page is atomically written, and it is in inmem_pages list. | |
203 | */ | |
d48dfc20 | 204 | #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1) |
0a595eba | 205 | #define DUMMY_WRITTEN_PAGE ((unsigned long)-2) |
decd36b6 CY |
206 | |
207 | #define IS_ATOMIC_WRITTEN_PAGE(page) \ | |
208 | (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE) | |
0a595eba JK |
209 | #define IS_DUMMY_WRITTEN_PAGE(page) \ |
210 | (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE) | |
decd36b6 | 211 | |
88b88a66 JK |
212 | struct inmem_pages { |
213 | struct list_head list; | |
214 | struct page *page; | |
28bc106b | 215 | block_t old_addr; /* for revoking when fail to commit */ |
88b88a66 JK |
216 | }; |
217 | ||
39a53e0c JK |
218 | struct sit_info { |
219 | const struct segment_allocation *s_ops; | |
220 | ||
221 | block_t sit_base_addr; /* start block address of SIT area */ | |
222 | block_t sit_blocks; /* # of blocks used by SIT area */ | |
223 | block_t written_valid_blocks; /* # of valid blocks in main area */ | |
224 | char *sit_bitmap; /* SIT bitmap pointer */ | |
ae27d62e CY |
225 | #ifdef CONFIG_F2FS_CHECK_FS |
226 | char *sit_bitmap_mir; /* SIT bitmap mirror */ | |
227 | #endif | |
39a53e0c JK |
228 | unsigned int bitmap_size; /* SIT bitmap size */ |
229 | ||
60a3b782 | 230 | unsigned long *tmp_map; /* bitmap for temporal use */ |
39a53e0c JK |
231 | unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ |
232 | unsigned int dirty_sentries; /* # of dirty sentries */ | |
233 | unsigned int sents_per_block; /* # of SIT entries per block */ | |
234 | struct mutex sentry_lock; /* to protect SIT cache */ | |
235 | struct seg_entry *sentries; /* SIT segment-level cache */ | |
236 | struct sec_entry *sec_entries; /* SIT section-level cache */ | |
237 | ||
238 | /* for cost-benefit algorithm in cleaning procedure */ | |
239 | unsigned long long elapsed_time; /* elapsed time after mount */ | |
240 | unsigned long long mounted_time; /* mount time */ | |
241 | unsigned long long min_mtime; /* min. modification time */ | |
242 | unsigned long long max_mtime; /* max. modification time */ | |
e066b83c JK |
243 | |
244 | unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */ | |
39a53e0c JK |
245 | }; |
246 | ||
247 | struct free_segmap_info { | |
248 | unsigned int start_segno; /* start segment number logically */ | |
249 | unsigned int free_segments; /* # of free segments */ | |
250 | unsigned int free_sections; /* # of free sections */ | |
1a118ccf | 251 | spinlock_t segmap_lock; /* free segmap lock */ |
39a53e0c JK |
252 | unsigned long *free_segmap; /* free segment bitmap */ |
253 | unsigned long *free_secmap; /* free section bitmap */ | |
254 | }; | |
255 | ||
256 | /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ | |
257 | enum dirty_type { | |
258 | DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ | |
259 | DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ | |
260 | DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ | |
261 | DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ | |
262 | DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ | |
263 | DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ | |
264 | DIRTY, /* to count # of dirty segments */ | |
265 | PRE, /* to count # of entirely obsolete segments */ | |
266 | NR_DIRTY_TYPE | |
267 | }; | |
268 | ||
269 | struct dirty_seglist_info { | |
270 | const struct victim_selection *v_ops; /* victim selction operation */ | |
271 | unsigned long *dirty_segmap[NR_DIRTY_TYPE]; | |
272 | struct mutex seglist_lock; /* lock for segment bitmaps */ | |
273 | int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ | |
5ec4e49f | 274 | unsigned long *victim_secmap; /* background GC victims */ |
39a53e0c JK |
275 | }; |
276 | ||
277 | /* victim selection function for cleaning and SSR */ | |
278 | struct victim_selection { | |
279 | int (*get_victim)(struct f2fs_sb_info *, unsigned int *, | |
280 | int, int, char); | |
281 | }; | |
282 | ||
283 | /* for active log information */ | |
284 | struct curseg_info { | |
285 | struct mutex curseg_mutex; /* lock for consistency */ | |
286 | struct f2fs_summary_block *sum_blk; /* cached summary block */ | |
b7ad7512 CY |
287 | struct rw_semaphore journal_rwsem; /* protect journal area */ |
288 | struct f2fs_journal *journal; /* cached journal info */ | |
39a53e0c JK |
289 | unsigned char alloc_type; /* current allocation type */ |
290 | unsigned int segno; /* current segment number */ | |
291 | unsigned short next_blkoff; /* next block offset to write */ | |
292 | unsigned int zone; /* current zone number */ | |
293 | unsigned int next_segno; /* preallocated segment */ | |
294 | }; | |
295 | ||
184a5cd2 CY |
296 | struct sit_entry_set { |
297 | struct list_head set_list; /* link with all sit sets */ | |
298 | unsigned int start_segno; /* start segno of sits in set */ | |
299 | unsigned int entry_cnt; /* the # of sit entries in set */ | |
300 | }; | |
301 | ||
39a53e0c JK |
302 | /* |
303 | * inline functions | |
304 | */ | |
305 | static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) | |
306 | { | |
307 | return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); | |
308 | } | |
309 | ||
310 | static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, | |
311 | unsigned int segno) | |
312 | { | |
313 | struct sit_info *sit_i = SIT_I(sbi); | |
314 | return &sit_i->sentries[segno]; | |
315 | } | |
316 | ||
317 | static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, | |
318 | unsigned int segno) | |
319 | { | |
320 | struct sit_info *sit_i = SIT_I(sbi); | |
4ddb1a4d | 321 | return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; |
39a53e0c JK |
322 | } |
323 | ||
324 | static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, | |
302bd348 | 325 | unsigned int segno, bool use_section) |
39a53e0c JK |
326 | { |
327 | /* | |
328 | * In order to get # of valid blocks in a section instantly from many | |
329 | * segments, f2fs manages two counting structures separately. | |
330 | */ | |
302bd348 | 331 | if (use_section && sbi->segs_per_sec > 1) |
39a53e0c JK |
332 | return get_sec_entry(sbi, segno)->valid_blocks; |
333 | else | |
334 | return get_seg_entry(sbi, segno)->valid_blocks; | |
335 | } | |
336 | ||
337 | static inline void seg_info_from_raw_sit(struct seg_entry *se, | |
338 | struct f2fs_sit_entry *rs) | |
339 | { | |
340 | se->valid_blocks = GET_SIT_VBLOCKS(rs); | |
341 | se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); | |
342 | memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
343 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
355e7891 CY |
344 | #ifdef CONFIG_F2FS_CHECK_FS |
345 | memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
346 | #endif | |
39a53e0c JK |
347 | se->type = GET_SIT_TYPE(rs); |
348 | se->mtime = le64_to_cpu(rs->mtime); | |
349 | } | |
350 | ||
351 | static inline void seg_info_to_raw_sit(struct seg_entry *se, | |
352 | struct f2fs_sit_entry *rs) | |
353 | { | |
354 | unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | | |
355 | se->valid_blocks; | |
356 | rs->vblocks = cpu_to_le16(raw_vblocks); | |
357 | memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); | |
358 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); | |
359 | se->ckpt_valid_blocks = se->valid_blocks; | |
360 | rs->mtime = cpu_to_le64(se->mtime); | |
361 | } | |
362 | ||
363 | static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, | |
364 | unsigned int max, unsigned int segno) | |
365 | { | |
366 | unsigned int ret; | |
1a118ccf | 367 | spin_lock(&free_i->segmap_lock); |
39a53e0c | 368 | ret = find_next_bit(free_i->free_segmap, max, segno); |
1a118ccf | 369 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
370 | return ret; |
371 | } | |
372 | ||
373 | static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) | |
374 | { | |
375 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
376 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
377 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
378 | unsigned int next; |
379 | ||
1a118ccf | 380 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
381 | clear_bit(segno, free_i->free_segmap); |
382 | free_i->free_segments++; | |
383 | ||
7fd97019 WL |
384 | next = find_next_bit(free_i->free_segmap, |
385 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
386 | if (next >= start_segno + sbi->segs_per_sec) { |
387 | clear_bit(secno, free_i->free_secmap); | |
388 | free_i->free_sections++; | |
389 | } | |
1a118ccf | 390 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
391 | } |
392 | ||
393 | static inline void __set_inuse(struct f2fs_sb_info *sbi, | |
394 | unsigned int segno) | |
395 | { | |
396 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
397 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
398 | ||
39a53e0c JK |
399 | set_bit(segno, free_i->free_segmap); |
400 | free_i->free_segments--; | |
401 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
402 | free_i->free_sections--; | |
403 | } | |
404 | ||
405 | static inline void __set_test_and_free(struct f2fs_sb_info *sbi, | |
406 | unsigned int segno) | |
407 | { | |
408 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
409 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
410 | unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); | |
39a53e0c JK |
411 | unsigned int next; |
412 | ||
1a118ccf | 413 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
414 | if (test_and_clear_bit(segno, free_i->free_segmap)) { |
415 | free_i->free_segments++; | |
416 | ||
f1121ab0 CY |
417 | next = find_next_bit(free_i->free_segmap, |
418 | start_segno + sbi->segs_per_sec, start_segno); | |
39a53e0c JK |
419 | if (next >= start_segno + sbi->segs_per_sec) { |
420 | if (test_and_clear_bit(secno, free_i->free_secmap)) | |
421 | free_i->free_sections++; | |
422 | } | |
423 | } | |
1a118ccf | 424 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
425 | } |
426 | ||
427 | static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, | |
428 | unsigned int segno) | |
429 | { | |
430 | struct free_segmap_info *free_i = FREE_I(sbi); | |
4ddb1a4d JK |
431 | unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); |
432 | ||
1a118ccf | 433 | spin_lock(&free_i->segmap_lock); |
39a53e0c JK |
434 | if (!test_and_set_bit(segno, free_i->free_segmap)) { |
435 | free_i->free_segments--; | |
436 | if (!test_and_set_bit(secno, free_i->free_secmap)) | |
437 | free_i->free_sections--; | |
438 | } | |
1a118ccf | 439 | spin_unlock(&free_i->segmap_lock); |
39a53e0c JK |
440 | } |
441 | ||
442 | static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, | |
443 | void *dst_addr) | |
444 | { | |
445 | struct sit_info *sit_i = SIT_I(sbi); | |
ae27d62e CY |
446 | |
447 | #ifdef CONFIG_F2FS_CHECK_FS | |
448 | if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, | |
449 | sit_i->bitmap_size)) | |
450 | f2fs_bug_on(sbi, 1); | |
451 | #endif | |
39a53e0c JK |
452 | memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); |
453 | } | |
454 | ||
455 | static inline block_t written_block_count(struct f2fs_sb_info *sbi) | |
456 | { | |
8b8343fa | 457 | return SIT_I(sbi)->written_valid_blocks; |
39a53e0c JK |
458 | } |
459 | ||
460 | static inline unsigned int free_segments(struct f2fs_sb_info *sbi) | |
461 | { | |
8b8343fa | 462 | return FREE_I(sbi)->free_segments; |
39a53e0c JK |
463 | } |
464 | ||
465 | static inline int reserved_segments(struct f2fs_sb_info *sbi) | |
466 | { | |
467 | return SM_I(sbi)->reserved_segments; | |
468 | } | |
469 | ||
470 | static inline unsigned int free_sections(struct f2fs_sb_info *sbi) | |
471 | { | |
8b8343fa | 472 | return FREE_I(sbi)->free_sections; |
39a53e0c JK |
473 | } |
474 | ||
475 | static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) | |
476 | { | |
477 | return DIRTY_I(sbi)->nr_dirty[PRE]; | |
478 | } | |
479 | ||
480 | static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) | |
481 | { | |
482 | return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + | |
483 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + | |
484 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + | |
485 | DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + | |
486 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + | |
487 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; | |
488 | } | |
489 | ||
490 | static inline int overprovision_segments(struct f2fs_sb_info *sbi) | |
491 | { | |
492 | return SM_I(sbi)->ovp_segments; | |
493 | } | |
494 | ||
39a53e0c JK |
495 | static inline int reserved_sections(struct f2fs_sb_info *sbi) |
496 | { | |
4ddb1a4d | 497 | return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi)); |
39a53e0c JK |
498 | } |
499 | ||
7f3037a5 JK |
500 | static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, |
501 | int freed, int needed) | |
39a53e0c | 502 | { |
5ac206cf NJ |
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); |
0f18b462 | 506 | |
caf0047e | 507 | if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) |
029cd28c JK |
508 | return false; |
509 | ||
7f3037a5 | 510 | return (free_sections(sbi) + freed) <= |
b9610bdf JK |
511 | (node_secs + 2 * dent_secs + imeta_secs + |
512 | reserved_sections(sbi) + needed); | |
39a53e0c JK |
513 | } |
514 | ||
81eb8d6e JK |
515 | static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) |
516 | { | |
6c311ec6 | 517 | return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; |
81eb8d6e JK |
518 | } |
519 | ||
39a53e0c JK |
520 | static inline int utilization(struct f2fs_sb_info *sbi) |
521 | { | |
6c311ec6 CF |
522 | return div_u64((u64)valid_user_blocks(sbi) * 100, |
523 | sbi->user_block_count); | |
39a53e0c JK |
524 | } |
525 | ||
526 | /* | |
527 | * Sometimes f2fs may be better to drop out-of-place update policy. | |
216fbd64 JK |
528 | * And, users can control the policy through sysfs entries. |
529 | * There are five policies with triggering conditions as follows. | |
530 | * F2FS_IPU_FORCE - all the time, | |
531 | * F2FS_IPU_SSR - if SSR mode is activated, | |
532 | * F2FS_IPU_UTIL - if FS utilization is over threashold, | |
533 | * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over | |
534 | * threashold, | |
c1ce1b02 JK |
535 | * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash |
536 | * storages. IPU will be triggered only if the # of dirty | |
537 | * pages over min_fsync_blocks. | |
216fbd64 | 538 | * F2FS_IPUT_DISABLE - disable IPU. (=default option) |
39a53e0c | 539 | */ |
216fbd64 | 540 | #define DEF_MIN_IPU_UTIL 70 |
c1ce1b02 | 541 | #define DEF_MIN_FSYNC_BLOCKS 8 |
ef095d19 | 542 | #define DEF_MIN_HOT_BLOCKS 16 |
216fbd64 JK |
543 | |
544 | enum { | |
545 | F2FS_IPU_FORCE, | |
546 | F2FS_IPU_SSR, | |
547 | F2FS_IPU_UTIL, | |
548 | F2FS_IPU_SSR_UTIL, | |
c1ce1b02 | 549 | F2FS_IPU_FSYNC, |
04485987 | 550 | F2FS_IPU_ASYNC, |
216fbd64 JK |
551 | }; |
552 | ||
7eab0c0d | 553 | static inline bool need_inplace_update_policy(struct inode *inode, |
04485987 | 554 | struct f2fs_io_info *fio) |
39a53e0c | 555 | { |
4081363f | 556 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); |
9b5f136f | 557 | unsigned int policy = SM_I(sbi)->ipu_policy; |
216fbd64 | 558 | |
36abef4e JK |
559 | if (test_opt(sbi, LFS)) |
560 | return false; | |
561 | ||
3537581a JK |
562 | /* if this is cold file, we should overwrite to avoid fragmentation */ |
563 | if (file_is_cold(inode)) | |
564 | return true; | |
565 | ||
9b5f136f | 566 | if (policy & (0x1 << F2FS_IPU_FORCE)) |
39a53e0c | 567 | return true; |
9b5f136f JK |
568 | if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi)) |
569 | return true; | |
570 | if (policy & (0x1 << F2FS_IPU_UTIL) && | |
571 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
572 | return true; | |
573 | if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) && | |
574 | utilization(sbi) > SM_I(sbi)->min_ipu_util) | |
575 | return true; | |
576 | ||
04485987 HP |
577 | /* |
578 | * IPU for rewrite async pages | |
579 | */ | |
580 | if (policy & (0x1 << F2FS_IPU_ASYNC) && | |
581 | fio && fio->op == REQ_OP_WRITE && | |
4086d3f6 HP |
582 | !(fio->op_flags & REQ_SYNC) && |
583 | !f2fs_encrypted_inode(inode)) | |
04485987 HP |
584 | return true; |
585 | ||
9b5f136f JK |
586 | /* this is only set during fdatasync */ |
587 | if (policy & (0x1 << F2FS_IPU_FSYNC) && | |
91942321 | 588 | is_inode_flag_set(inode, FI_NEED_IPU)) |
9b5f136f JK |
589 | return true; |
590 | ||
39a53e0c JK |
591 | return false; |
592 | } | |
593 | ||
594 | static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, | |
595 | int type) | |
596 | { | |
597 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
598 | return curseg->segno; | |
599 | } | |
600 | ||
601 | static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, | |
602 | int type) | |
603 | { | |
604 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
605 | return curseg->alloc_type; | |
606 | } | |
607 | ||
608 | static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type) | |
609 | { | |
610 | struct curseg_info *curseg = CURSEG_I(sbi, type); | |
611 | return curseg->next_blkoff; | |
612 | } | |
613 | ||
614 | static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) | |
615 | { | |
7a04f64d | 616 | f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1); |
39a53e0c JK |
617 | } |
618 | ||
39a53e0c JK |
619 | static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr) |
620 | { | |
bb413d6a YH |
621 | BUG_ON(blk_addr < SEG0_BLKADDR(sbi) |
622 | || blk_addr >= MAX_BLKADDR(sbi)); | |
39a53e0c JK |
623 | } |
624 | ||
625 | /* | |
e1c42045 | 626 | * Summary block is always treated as an invalid block |
39a53e0c JK |
627 | */ |
628 | static inline void check_block_count(struct f2fs_sb_info *sbi, | |
629 | int segno, struct f2fs_sit_entry *raw_sit) | |
630 | { | |
4c278394 | 631 | #ifdef CONFIG_F2FS_CHECK_FS |
44c60bf2 | 632 | bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; |
39a53e0c | 633 | int valid_blocks = 0; |
44c60bf2 | 634 | int cur_pos = 0, next_pos; |
39a53e0c | 635 | |
39a53e0c | 636 | /* check bitmap with valid block count */ |
44c60bf2 CY |
637 | do { |
638 | if (is_valid) { | |
639 | next_pos = find_next_zero_bit_le(&raw_sit->valid_map, | |
640 | sbi->blocks_per_seg, | |
641 | cur_pos); | |
642 | valid_blocks += next_pos - cur_pos; | |
643 | } else | |
644 | next_pos = find_next_bit_le(&raw_sit->valid_map, | |
645 | sbi->blocks_per_seg, | |
646 | cur_pos); | |
647 | cur_pos = next_pos; | |
648 | is_valid = !is_valid; | |
649 | } while (cur_pos < sbi->blocks_per_seg); | |
39a53e0c | 650 | BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks); |
5d56b671 | 651 | #endif |
4c278394 JK |
652 | /* check segment usage, and check boundary of a given segment number */ |
653 | f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg | |
654 | || segno > TOTAL_SEGS(sbi) - 1); | |
7a04f64d | 655 | } |
39a53e0c JK |
656 | |
657 | static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, | |
658 | unsigned int start) | |
659 | { | |
660 | struct sit_info *sit_i = SIT_I(sbi); | |
d3a14afd | 661 | unsigned int offset = SIT_BLOCK_OFFSET(start); |
39a53e0c JK |
662 | block_t blk_addr = sit_i->sit_base_addr + offset; |
663 | ||
664 | check_seg_range(sbi, start); | |
665 | ||
ae27d62e CY |
666 | #ifdef CONFIG_F2FS_CHECK_FS |
667 | if (f2fs_test_bit(offset, sit_i->sit_bitmap) != | |
668 | f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) | |
669 | f2fs_bug_on(sbi, 1); | |
670 | #endif | |
671 | ||
39a53e0c JK |
672 | /* calculate sit block address */ |
673 | if (f2fs_test_bit(offset, sit_i->sit_bitmap)) | |
674 | blk_addr += sit_i->sit_blocks; | |
675 | ||
676 | return blk_addr; | |
677 | } | |
678 | ||
679 | static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, | |
680 | pgoff_t block_addr) | |
681 | { | |
682 | struct sit_info *sit_i = SIT_I(sbi); | |
683 | block_addr -= sit_i->sit_base_addr; | |
684 | if (block_addr < sit_i->sit_blocks) | |
685 | block_addr += sit_i->sit_blocks; | |
686 | else | |
687 | block_addr -= sit_i->sit_blocks; | |
688 | ||
689 | return block_addr + sit_i->sit_base_addr; | |
690 | } | |
691 | ||
692 | static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) | |
693 | { | |
d3a14afd | 694 | unsigned int block_off = SIT_BLOCK_OFFSET(start); |
39a53e0c | 695 | |
c6ac4c0e | 696 | f2fs_change_bit(block_off, sit_i->sit_bitmap); |
ae27d62e CY |
697 | #ifdef CONFIG_F2FS_CHECK_FS |
698 | f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); | |
699 | #endif | |
39a53e0c JK |
700 | } |
701 | ||
702 | static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi) | |
703 | { | |
704 | struct sit_info *sit_i = SIT_I(sbi); | |
48fbfe50 DD |
705 | time64_t now = ktime_get_real_seconds(); |
706 | ||
707 | return sit_i->elapsed_time + now - sit_i->mounted_time; | |
39a53e0c JK |
708 | } |
709 | ||
710 | static inline void set_summary(struct f2fs_summary *sum, nid_t nid, | |
711 | unsigned int ofs_in_node, unsigned char version) | |
712 | { | |
713 | sum->nid = cpu_to_le32(nid); | |
714 | sum->ofs_in_node = cpu_to_le16(ofs_in_node); | |
715 | sum->version = version; | |
716 | } | |
717 | ||
718 | static inline block_t start_sum_block(struct f2fs_sb_info *sbi) | |
719 | { | |
720 | return __start_cp_addr(sbi) + | |
721 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); | |
722 | } | |
723 | ||
724 | static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) | |
725 | { | |
726 | return __start_cp_addr(sbi) + | |
727 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) | |
728 | - (base + 1) + type; | |
729 | } | |
5ec4e49f | 730 | |
e93b9865 HP |
731 | static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi, |
732 | unsigned int secno) | |
733 | { | |
4ddb1a4d | 734 | if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >= |
302bd348 | 735 | sbi->fggc_threshold) |
e93b9865 HP |
736 | return true; |
737 | return false; | |
738 | } | |
739 | ||
5ec4e49f JK |
740 | static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) |
741 | { | |
742 | if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) | |
743 | return true; | |
744 | return false; | |
745 | } | |
ac5d156c | 746 | |
87d6f890 JK |
747 | /* |
748 | * It is very important to gather dirty pages and write at once, so that we can | |
749 | * submit a big bio without interfering other data writes. | |
750 | * By default, 512 pages for directory data, | |
727ebb09 KM |
751 | * 512 pages (2MB) * 8 for nodes, and |
752 | * 256 pages * 8 for meta are set. | |
87d6f890 JK |
753 | */ |
754 | static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) | |
755 | { | |
a88a341a | 756 | if (sbi->sb->s_bdi->wb.dirty_exceeded) |
510184c8 JK |
757 | return 0; |
758 | ||
a1257023 JK |
759 | if (type == DATA) |
760 | return sbi->blocks_per_seg; | |
761 | else if (type == NODE) | |
2c237eba | 762 | return 8 * sbi->blocks_per_seg; |
87d6f890 | 763 | else if (type == META) |
664ba972 | 764 | return 8 * BIO_MAX_PAGES; |
87d6f890 JK |
765 | else |
766 | return 0; | |
767 | } | |
50c8cdb3 JK |
768 | |
769 | /* | |
770 | * When writing pages, it'd better align nr_to_write for segment size. | |
771 | */ | |
772 | static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, | |
773 | struct writeback_control *wbc) | |
774 | { | |
775 | long nr_to_write, desired; | |
776 | ||
777 | if (wbc->sync_mode != WB_SYNC_NONE) | |
778 | return 0; | |
779 | ||
780 | nr_to_write = wbc->nr_to_write; | |
664ba972 | 781 | desired = BIO_MAX_PAGES; |
28ea6162 | 782 | if (type == NODE) |
664ba972 | 783 | desired <<= 1; |
50c8cdb3 JK |
784 | |
785 | wbc->nr_to_write = desired; | |
786 | return desired - nr_to_write; | |
787 | } | |
01983c71 JK |
788 | |
789 | static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force) | |
790 | { | |
791 | struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info; | |
792 | bool wakeup = false; | |
793 | int i; | |
794 | ||
795 | if (force) | |
796 | goto wake_up; | |
797 | ||
798 | mutex_lock(&dcc->cmd_lock); | |
799 | for (i = MAX_PLIST_NUM - 1; | |
800 | i >= 0 && plist_issue(dcc->pend_list_tag[i]); i--) { | |
801 | if (!list_empty(&dcc->pend_list[i])) { | |
802 | wakeup = true; | |
803 | break; | |
804 | } | |
805 | } | |
806 | mutex_unlock(&dcc->cmd_lock); | |
807 | if (!wakeup) | |
808 | return; | |
809 | wake_up: | |
810 | dcc->discard_wake = 1; | |
811 | wake_up_interruptible_all(&dcc->discard_wait_queue); | |
812 | } |