Jaegeuk Kim | 0a8165d | 2012-11-29 13:28:09 +0900 | [diff] [blame] | 1 | /* |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 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 | */ |
Jaegeuk Kim | ac5d156 | 2013-04-29 16:58:39 +0900 | [diff] [blame] | 11 | #include <linux/blkdev.h> |
| 12 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 13 | /* constant macro */ |
| 14 | #define NULL_SEGNO ((unsigned int)(~0)) |
Jaegeuk Kim | 5ec4e49 | 2013-03-31 13:26:03 +0900 | [diff] [blame] | 15 | #define NULL_SECNO ((unsigned int)(~0)) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 16 | |
Jaegeuk Kim | 58c4103 | 2014-03-19 14:17:21 +0900 | [diff] [blame] | 17 | #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */ |
Jaegeuk Kim | 81eb8d6 | 2013-10-24 13:31:34 +0900 | [diff] [blame] | 18 | |
Namjae Jeon | 6224da8 | 2013-04-06 14:44:32 +0900 | [diff] [blame] | 19 | /* L: Logical segment # in volume, R: Relative segment # in main area */ |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 20 | #define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno) |
| 21 | #define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno) |
| 22 | |
Changman Lee | 61ae45c | 2013-11-21 20:04:21 +0900 | [diff] [blame] | 23 | #define IS_DATASEG(t) (t <= CURSEG_COLD_DATA) |
| 24 | #define IS_NODESEG(t) (t >= CURSEG_HOT_NODE) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 25 | |
Jaegeuk Kim | 5c773ba | 2013-03-31 12:30:04 +0900 | [diff] [blame] | 26 | #define IS_CURSEG(sbi, seg) \ |
| 27 | ((seg == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \ |
| 28 | (seg == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \ |
| 29 | (seg == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \ |
| 30 | (seg == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \ |
| 31 | (seg == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \ |
| 32 | (seg == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno)) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 33 | |
| 34 | #define IS_CURSEC(sbi, secno) \ |
| 35 | ((secno == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \ |
| 36 | sbi->segs_per_sec) || \ |
| 37 | (secno == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \ |
| 38 | sbi->segs_per_sec) || \ |
| 39 | (secno == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \ |
| 40 | sbi->segs_per_sec) || \ |
| 41 | (secno == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \ |
| 42 | sbi->segs_per_sec) || \ |
| 43 | (secno == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \ |
| 44 | sbi->segs_per_sec) || \ |
| 45 | (secno == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \ |
| 46 | sbi->segs_per_sec)) \ |
| 47 | |
| 48 | #define START_BLOCK(sbi, segno) \ |
| 49 | (SM_I(sbi)->seg0_blkaddr + \ |
| 50 | (GET_R2L_SEGNO(FREE_I(sbi), segno) << sbi->log_blocks_per_seg)) |
| 51 | #define NEXT_FREE_BLKADDR(sbi, curseg) \ |
| 52 | (START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff) |
| 53 | |
| 54 | #define MAIN_BASE_BLOCK(sbi) (SM_I(sbi)->main_blkaddr) |
| 55 | |
| 56 | #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) \ |
| 57 | ((blk_addr) - SM_I(sbi)->seg0_blkaddr) |
| 58 | #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ |
| 59 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> sbi->log_blocks_per_seg) |
Jaegeuk Kim | 491c085 | 2014-02-04 13:01:10 +0900 | [diff] [blame] | 60 | #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ |
| 61 | (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (sbi->blocks_per_seg - 1)) |
| 62 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 63 | #define GET_SEGNO(sbi, blk_addr) \ |
| 64 | (((blk_addr == NULL_ADDR) || (blk_addr == NEW_ADDR)) ? \ |
| 65 | NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ |
| 66 | GET_SEGNO_FROM_SEG0(sbi, blk_addr))) |
| 67 | #define GET_SECNO(sbi, segno) \ |
| 68 | ((segno) / sbi->segs_per_sec) |
| 69 | #define GET_ZONENO_FROM_SEGNO(sbi, segno) \ |
| 70 | ((segno / sbi->segs_per_sec) / sbi->secs_per_zone) |
| 71 | |
| 72 | #define GET_SUM_BLOCK(sbi, segno) \ |
| 73 | ((sbi->sm_info->ssa_blkaddr) + segno) |
| 74 | |
| 75 | #define GET_SUM_TYPE(footer) ((footer)->entry_type) |
| 76 | #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = type) |
| 77 | |
| 78 | #define SIT_ENTRY_OFFSET(sit_i, segno) \ |
| 79 | (segno % sit_i->sents_per_block) |
| 80 | #define SIT_BLOCK_OFFSET(sit_i, segno) \ |
| 81 | (segno / SIT_ENTRY_PER_BLOCK) |
| 82 | #define START_SEGNO(sit_i, segno) \ |
| 83 | (SIT_BLOCK_OFFSET(sit_i, segno) * SIT_ENTRY_PER_BLOCK) |
Chao Yu | 74de593 | 2013-11-22 09:09:59 +0800 | [diff] [blame] | 84 | #define SIT_BLK_CNT(sbi) \ |
| 85 | ((TOTAL_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 86 | #define f2fs_bitmap_size(nr) \ |
| 87 | (BITS_TO_LONGS(nr) * sizeof(unsigned long)) |
| 88 | #define TOTAL_SEGS(sbi) (SM_I(sbi)->main_segments) |
Jaegeuk Kim | 53cf952 | 2013-03-31 12:39:49 +0900 | [diff] [blame] | 89 | #define TOTAL_SECS(sbi) (sbi->total_sections) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 90 | |
Jaegeuk Kim | 3cd8a23 | 2012-12-10 09:26:05 +0900 | [diff] [blame] | 91 | #define SECTOR_FROM_BLOCK(sbi, blk_addr) \ |
Jaegeuk Kim | f9a4e6d | 2013-11-28 12:44:05 +0900 | [diff] [blame] | 92 | (((sector_t)blk_addr) << (sbi)->log_sectors_per_block) |
Jaegeuk Kim | ac5d156 | 2013-04-29 16:58:39 +0900 | [diff] [blame] | 93 | #define SECTOR_TO_BLOCK(sbi, sectors) \ |
Jaegeuk Kim | f9a4e6d | 2013-11-28 12:44:05 +0900 | [diff] [blame] | 94 | (sectors >> (sbi)->log_sectors_per_block) |
Chao Yu | cc7b1bb | 2013-09-22 15:50:50 +0800 | [diff] [blame] | 95 | #define MAX_BIO_BLOCKS(max_hw_blocks) \ |
| 96 | (min((int)max_hw_blocks, BIO_MAX_PAGES)) |
Jaegeuk Kim | 3cd8a23 | 2012-12-10 09:26:05 +0900 | [diff] [blame] | 97 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 98 | /* |
| 99 | * indicate a block allocation direction: RIGHT and LEFT. |
| 100 | * RIGHT means allocating new sections towards the end of volume. |
| 101 | * LEFT means the opposite direction. |
| 102 | */ |
| 103 | enum { |
| 104 | ALLOC_RIGHT = 0, |
| 105 | ALLOC_LEFT |
| 106 | }; |
| 107 | |
| 108 | /* |
| 109 | * In the victim_sel_policy->alloc_mode, there are two block allocation modes. |
| 110 | * LFS writes data sequentially with cleaning operations. |
| 111 | * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. |
| 112 | */ |
| 113 | enum { |
| 114 | LFS = 0, |
| 115 | SSR |
| 116 | }; |
| 117 | |
| 118 | /* |
| 119 | * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes. |
| 120 | * GC_CB is based on cost-benefit algorithm. |
| 121 | * GC_GREEDY is based on greedy algorithm. |
| 122 | */ |
| 123 | enum { |
| 124 | GC_CB = 0, |
| 125 | GC_GREEDY |
| 126 | }; |
| 127 | |
| 128 | /* |
| 129 | * BG_GC means the background cleaning job. |
| 130 | * FG_GC means the on-demand cleaning job. |
| 131 | */ |
| 132 | enum { |
| 133 | BG_GC = 0, |
| 134 | FG_GC |
| 135 | }; |
| 136 | |
| 137 | /* for a function parameter to select a victim segment */ |
| 138 | struct victim_sel_policy { |
| 139 | int alloc_mode; /* LFS or SSR */ |
| 140 | int gc_mode; /* GC_CB or GC_GREEDY */ |
| 141 | unsigned long *dirty_segmap; /* dirty segment bitmap */ |
Jin Xu | a26b7c8 | 2013-09-05 12:45:26 +0800 | [diff] [blame] | 142 | unsigned int max_search; /* maximum # of segments to search */ |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 143 | unsigned int offset; /* last scanned bitmap offset */ |
| 144 | unsigned int ofs_unit; /* bitmap search unit */ |
| 145 | unsigned int min_cost; /* minimum cost */ |
| 146 | unsigned int min_segno; /* segment # having min. cost */ |
| 147 | }; |
| 148 | |
| 149 | struct seg_entry { |
| 150 | unsigned short valid_blocks; /* # of valid blocks */ |
| 151 | unsigned char *cur_valid_map; /* validity bitmap of blocks */ |
| 152 | /* |
| 153 | * # of valid blocks and the validity bitmap stored in the the last |
| 154 | * checkpoint pack. This information is used by the SSR mode. |
| 155 | */ |
| 156 | unsigned short ckpt_valid_blocks; |
| 157 | unsigned char *ckpt_valid_map; |
| 158 | unsigned char type; /* segment type like CURSEG_XXX_TYPE */ |
| 159 | unsigned long long mtime; /* modification time of the segment */ |
| 160 | }; |
| 161 | |
| 162 | struct sec_entry { |
| 163 | unsigned int valid_blocks; /* # of valid blocks in a section */ |
| 164 | }; |
| 165 | |
| 166 | struct segment_allocation { |
| 167 | void (*allocate_segment)(struct f2fs_sb_info *, int, bool); |
| 168 | }; |
| 169 | |
| 170 | struct sit_info { |
| 171 | const struct segment_allocation *s_ops; |
| 172 | |
| 173 | block_t sit_base_addr; /* start block address of SIT area */ |
| 174 | block_t sit_blocks; /* # of blocks used by SIT area */ |
| 175 | block_t written_valid_blocks; /* # of valid blocks in main area */ |
| 176 | char *sit_bitmap; /* SIT bitmap pointer */ |
| 177 | unsigned int bitmap_size; /* SIT bitmap size */ |
| 178 | |
| 179 | unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ |
| 180 | unsigned int dirty_sentries; /* # of dirty sentries */ |
| 181 | unsigned int sents_per_block; /* # of SIT entries per block */ |
| 182 | struct mutex sentry_lock; /* to protect SIT cache */ |
| 183 | struct seg_entry *sentries; /* SIT segment-level cache */ |
| 184 | struct sec_entry *sec_entries; /* SIT section-level cache */ |
| 185 | |
| 186 | /* for cost-benefit algorithm in cleaning procedure */ |
| 187 | unsigned long long elapsed_time; /* elapsed time after mount */ |
| 188 | unsigned long long mounted_time; /* mount time */ |
| 189 | unsigned long long min_mtime; /* min. modification time */ |
| 190 | unsigned long long max_mtime; /* max. modification time */ |
| 191 | }; |
| 192 | |
| 193 | struct free_segmap_info { |
| 194 | unsigned int start_segno; /* start segment number logically */ |
| 195 | unsigned int free_segments; /* # of free segments */ |
| 196 | unsigned int free_sections; /* # of free sections */ |
| 197 | rwlock_t segmap_lock; /* free segmap lock */ |
| 198 | unsigned long *free_segmap; /* free segment bitmap */ |
| 199 | unsigned long *free_secmap; /* free section bitmap */ |
| 200 | }; |
| 201 | |
| 202 | /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ |
| 203 | enum dirty_type { |
| 204 | DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ |
| 205 | DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ |
| 206 | DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ |
| 207 | DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ |
| 208 | DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ |
| 209 | DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ |
| 210 | DIRTY, /* to count # of dirty segments */ |
| 211 | PRE, /* to count # of entirely obsolete segments */ |
| 212 | NR_DIRTY_TYPE |
| 213 | }; |
| 214 | |
| 215 | struct dirty_seglist_info { |
| 216 | const struct victim_selection *v_ops; /* victim selction operation */ |
| 217 | unsigned long *dirty_segmap[NR_DIRTY_TYPE]; |
| 218 | struct mutex seglist_lock; /* lock for segment bitmaps */ |
| 219 | int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ |
Jaegeuk Kim | 5ec4e49 | 2013-03-31 13:26:03 +0900 | [diff] [blame] | 220 | unsigned long *victim_secmap; /* background GC victims */ |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 221 | }; |
| 222 | |
| 223 | /* victim selection function for cleaning and SSR */ |
| 224 | struct victim_selection { |
| 225 | int (*get_victim)(struct f2fs_sb_info *, unsigned int *, |
| 226 | int, int, char); |
| 227 | }; |
| 228 | |
| 229 | /* for active log information */ |
| 230 | struct curseg_info { |
| 231 | struct mutex curseg_mutex; /* lock for consistency */ |
| 232 | struct f2fs_summary_block *sum_blk; /* cached summary block */ |
| 233 | unsigned char alloc_type; /* current allocation type */ |
| 234 | unsigned int segno; /* current segment number */ |
| 235 | unsigned short next_blkoff; /* next block offset to write */ |
| 236 | unsigned int zone; /* current zone number */ |
| 237 | unsigned int next_segno; /* preallocated segment */ |
| 238 | }; |
| 239 | |
| 240 | /* |
| 241 | * inline functions |
| 242 | */ |
| 243 | static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) |
| 244 | { |
| 245 | return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); |
| 246 | } |
| 247 | |
| 248 | static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, |
| 249 | unsigned int segno) |
| 250 | { |
| 251 | struct sit_info *sit_i = SIT_I(sbi); |
| 252 | return &sit_i->sentries[segno]; |
| 253 | } |
| 254 | |
| 255 | static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, |
| 256 | unsigned int segno) |
| 257 | { |
| 258 | struct sit_info *sit_i = SIT_I(sbi); |
| 259 | return &sit_i->sec_entries[GET_SECNO(sbi, segno)]; |
| 260 | } |
| 261 | |
| 262 | static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, |
| 263 | unsigned int segno, int section) |
| 264 | { |
| 265 | /* |
| 266 | * In order to get # of valid blocks in a section instantly from many |
| 267 | * segments, f2fs manages two counting structures separately. |
| 268 | */ |
| 269 | if (section > 1) |
| 270 | return get_sec_entry(sbi, segno)->valid_blocks; |
| 271 | else |
| 272 | return get_seg_entry(sbi, segno)->valid_blocks; |
| 273 | } |
| 274 | |
| 275 | static inline void seg_info_from_raw_sit(struct seg_entry *se, |
| 276 | struct f2fs_sit_entry *rs) |
| 277 | { |
| 278 | se->valid_blocks = GET_SIT_VBLOCKS(rs); |
| 279 | se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); |
| 280 | memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); |
| 281 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); |
| 282 | se->type = GET_SIT_TYPE(rs); |
| 283 | se->mtime = le64_to_cpu(rs->mtime); |
| 284 | } |
| 285 | |
| 286 | static inline void seg_info_to_raw_sit(struct seg_entry *se, |
| 287 | struct f2fs_sit_entry *rs) |
| 288 | { |
| 289 | unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | |
| 290 | se->valid_blocks; |
| 291 | rs->vblocks = cpu_to_le16(raw_vblocks); |
| 292 | memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); |
| 293 | memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); |
| 294 | se->ckpt_valid_blocks = se->valid_blocks; |
| 295 | rs->mtime = cpu_to_le64(se->mtime); |
| 296 | } |
| 297 | |
| 298 | static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, |
| 299 | unsigned int max, unsigned int segno) |
| 300 | { |
| 301 | unsigned int ret; |
| 302 | read_lock(&free_i->segmap_lock); |
| 303 | ret = find_next_bit(free_i->free_segmap, max, segno); |
| 304 | read_unlock(&free_i->segmap_lock); |
| 305 | return ret; |
| 306 | } |
| 307 | |
| 308 | static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) |
| 309 | { |
| 310 | struct free_segmap_info *free_i = FREE_I(sbi); |
| 311 | unsigned int secno = segno / sbi->segs_per_sec; |
| 312 | unsigned int start_segno = secno * sbi->segs_per_sec; |
| 313 | unsigned int next; |
| 314 | |
| 315 | write_lock(&free_i->segmap_lock); |
| 316 | clear_bit(segno, free_i->free_segmap); |
| 317 | free_i->free_segments++; |
| 318 | |
| 319 | next = find_next_bit(free_i->free_segmap, TOTAL_SEGS(sbi), start_segno); |
| 320 | if (next >= start_segno + sbi->segs_per_sec) { |
| 321 | clear_bit(secno, free_i->free_secmap); |
| 322 | free_i->free_sections++; |
| 323 | } |
| 324 | write_unlock(&free_i->segmap_lock); |
| 325 | } |
| 326 | |
| 327 | static inline void __set_inuse(struct f2fs_sb_info *sbi, |
| 328 | unsigned int segno) |
| 329 | { |
| 330 | struct free_segmap_info *free_i = FREE_I(sbi); |
| 331 | unsigned int secno = segno / sbi->segs_per_sec; |
| 332 | set_bit(segno, free_i->free_segmap); |
| 333 | free_i->free_segments--; |
| 334 | if (!test_and_set_bit(secno, free_i->free_secmap)) |
| 335 | free_i->free_sections--; |
| 336 | } |
| 337 | |
| 338 | static inline void __set_test_and_free(struct f2fs_sb_info *sbi, |
| 339 | unsigned int segno) |
| 340 | { |
| 341 | struct free_segmap_info *free_i = FREE_I(sbi); |
| 342 | unsigned int secno = segno / sbi->segs_per_sec; |
| 343 | unsigned int start_segno = secno * sbi->segs_per_sec; |
| 344 | unsigned int next; |
| 345 | |
| 346 | write_lock(&free_i->segmap_lock); |
| 347 | if (test_and_clear_bit(segno, free_i->free_segmap)) { |
| 348 | free_i->free_segments++; |
| 349 | |
Chao Yu | f1121ab | 2014-07-14 16:45:15 +0800 | [diff] [blame] | 350 | next = find_next_bit(free_i->free_segmap, |
| 351 | start_segno + sbi->segs_per_sec, start_segno); |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 352 | if (next >= start_segno + sbi->segs_per_sec) { |
| 353 | if (test_and_clear_bit(secno, free_i->free_secmap)) |
| 354 | free_i->free_sections++; |
| 355 | } |
| 356 | } |
| 357 | write_unlock(&free_i->segmap_lock); |
| 358 | } |
| 359 | |
| 360 | static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, |
| 361 | unsigned int segno) |
| 362 | { |
| 363 | struct free_segmap_info *free_i = FREE_I(sbi); |
| 364 | unsigned int secno = segno / sbi->segs_per_sec; |
| 365 | write_lock(&free_i->segmap_lock); |
| 366 | if (!test_and_set_bit(segno, free_i->free_segmap)) { |
| 367 | free_i->free_segments--; |
| 368 | if (!test_and_set_bit(secno, free_i->free_secmap)) |
| 369 | free_i->free_sections--; |
| 370 | } |
| 371 | write_unlock(&free_i->segmap_lock); |
| 372 | } |
| 373 | |
| 374 | static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, |
| 375 | void *dst_addr) |
| 376 | { |
| 377 | struct sit_info *sit_i = SIT_I(sbi); |
| 378 | memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); |
| 379 | } |
| 380 | |
| 381 | static inline block_t written_block_count(struct f2fs_sb_info *sbi) |
| 382 | { |
Jaegeuk Kim | 8b8343f | 2014-02-24 13:00:13 +0900 | [diff] [blame] | 383 | return SIT_I(sbi)->written_valid_blocks; |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 384 | } |
| 385 | |
| 386 | static inline unsigned int free_segments(struct f2fs_sb_info *sbi) |
| 387 | { |
Jaegeuk Kim | 8b8343f | 2014-02-24 13:00:13 +0900 | [diff] [blame] | 388 | return FREE_I(sbi)->free_segments; |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 389 | } |
| 390 | |
| 391 | static inline int reserved_segments(struct f2fs_sb_info *sbi) |
| 392 | { |
| 393 | return SM_I(sbi)->reserved_segments; |
| 394 | } |
| 395 | |
| 396 | static inline unsigned int free_sections(struct f2fs_sb_info *sbi) |
| 397 | { |
Jaegeuk Kim | 8b8343f | 2014-02-24 13:00:13 +0900 | [diff] [blame] | 398 | return FREE_I(sbi)->free_sections; |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 399 | } |
| 400 | |
| 401 | static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) |
| 402 | { |
| 403 | return DIRTY_I(sbi)->nr_dirty[PRE]; |
| 404 | } |
| 405 | |
| 406 | static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) |
| 407 | { |
| 408 | return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + |
| 409 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + |
| 410 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + |
| 411 | DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + |
| 412 | DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + |
| 413 | DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; |
| 414 | } |
| 415 | |
| 416 | static inline int overprovision_segments(struct f2fs_sb_info *sbi) |
| 417 | { |
| 418 | return SM_I(sbi)->ovp_segments; |
| 419 | } |
| 420 | |
| 421 | static inline int overprovision_sections(struct f2fs_sb_info *sbi) |
| 422 | { |
| 423 | return ((unsigned int) overprovision_segments(sbi)) / sbi->segs_per_sec; |
| 424 | } |
| 425 | |
| 426 | static inline int reserved_sections(struct f2fs_sb_info *sbi) |
| 427 | { |
| 428 | return ((unsigned int) reserved_segments(sbi)) / sbi->segs_per_sec; |
| 429 | } |
| 430 | |
| 431 | static inline bool need_SSR(struct f2fs_sb_info *sbi) |
| 432 | { |
Chris Fries | 6c311ec | 2014-01-17 14:44:39 -0600 | [diff] [blame] | 433 | return (prefree_segments(sbi) / sbi->segs_per_sec) |
| 434 | + free_sections(sbi) < overprovision_sections(sbi); |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 435 | } |
| 436 | |
Jaegeuk Kim | 4372752 | 2013-02-04 15:11:17 +0900 | [diff] [blame] | 437 | static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, int freed) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 438 | { |
Namjae Jeon | 5ac206c | 2013-02-02 23:52:59 +0900 | [diff] [blame] | 439 | int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); |
| 440 | int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); |
Jaegeuk Kim | 4372752 | 2013-02-04 15:11:17 +0900 | [diff] [blame] | 441 | |
Chao Yu | cfb271d | 2013-12-05 17:15:22 +0800 | [diff] [blame] | 442 | if (unlikely(sbi->por_doing)) |
Jaegeuk Kim | 029cd28 | 2012-12-21 17:20:21 +0900 | [diff] [blame] | 443 | return false; |
| 444 | |
Chris Fries | 6c311ec | 2014-01-17 14:44:39 -0600 | [diff] [blame] | 445 | return (free_sections(sbi) + freed) <= (node_secs + 2 * dent_secs + |
| 446 | reserved_sections(sbi)); |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 447 | } |
| 448 | |
Jaegeuk Kim | 81eb8d6 | 2013-10-24 13:31:34 +0900 | [diff] [blame] | 449 | static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) |
| 450 | { |
Chris Fries | 6c311ec | 2014-01-17 14:44:39 -0600 | [diff] [blame] | 451 | return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; |
Jaegeuk Kim | 81eb8d6 | 2013-10-24 13:31:34 +0900 | [diff] [blame] | 452 | } |
| 453 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 454 | static inline int utilization(struct f2fs_sb_info *sbi) |
| 455 | { |
Chris Fries | 6c311ec | 2014-01-17 14:44:39 -0600 | [diff] [blame] | 456 | return div_u64((u64)valid_user_blocks(sbi) * 100, |
| 457 | sbi->user_block_count); |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 458 | } |
| 459 | |
| 460 | /* |
| 461 | * Sometimes f2fs may be better to drop out-of-place update policy. |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 462 | * And, users can control the policy through sysfs entries. |
| 463 | * There are five policies with triggering conditions as follows. |
| 464 | * F2FS_IPU_FORCE - all the time, |
| 465 | * F2FS_IPU_SSR - if SSR mode is activated, |
| 466 | * F2FS_IPU_UTIL - if FS utilization is over threashold, |
| 467 | * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over |
| 468 | * threashold, |
| 469 | * F2FS_IPUT_DISABLE - disable IPU. (=default option) |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 470 | */ |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 471 | #define DEF_MIN_IPU_UTIL 70 |
| 472 | |
| 473 | enum { |
| 474 | F2FS_IPU_FORCE, |
| 475 | F2FS_IPU_SSR, |
| 476 | F2FS_IPU_UTIL, |
| 477 | F2FS_IPU_SSR_UTIL, |
| 478 | F2FS_IPU_DISABLE, |
| 479 | }; |
| 480 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 481 | static inline bool need_inplace_update(struct inode *inode) |
| 482 | { |
| 483 | struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 484 | |
| 485 | /* IPU can be done only for the user data */ |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 486 | if (S_ISDIR(inode->i_mode)) |
| 487 | return false; |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 488 | |
Jaegeuk Kim | ea1aa12 | 2014-07-24 19:11:43 -0700 | [diff] [blame] | 489 | /* this is only set during fdatasync */ |
| 490 | if (is_inode_flag_set(F2FS_I(inode), FI_NEED_IPU)) |
| 491 | return true; |
| 492 | |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 493 | switch (SM_I(sbi)->ipu_policy) { |
| 494 | case F2FS_IPU_FORCE: |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 495 | return true; |
Jaegeuk Kim | 216fbd6 | 2013-11-07 13:13:42 +0900 | [diff] [blame] | 496 | case F2FS_IPU_SSR: |
| 497 | if (need_SSR(sbi)) |
| 498 | return true; |
| 499 | break; |
| 500 | case F2FS_IPU_UTIL: |
| 501 | if (utilization(sbi) > SM_I(sbi)->min_ipu_util) |
| 502 | return true; |
| 503 | break; |
| 504 | case F2FS_IPU_SSR_UTIL: |
| 505 | if (need_SSR(sbi) && utilization(sbi) > SM_I(sbi)->min_ipu_util) |
| 506 | return true; |
| 507 | break; |
| 508 | case F2FS_IPU_DISABLE: |
| 509 | break; |
| 510 | } |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 511 | return false; |
| 512 | } |
| 513 | |
| 514 | static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, |
| 515 | int type) |
| 516 | { |
| 517 | struct curseg_info *curseg = CURSEG_I(sbi, type); |
| 518 | return curseg->segno; |
| 519 | } |
| 520 | |
| 521 | static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, |
| 522 | int type) |
| 523 | { |
| 524 | struct curseg_info *curseg = CURSEG_I(sbi, type); |
| 525 | return curseg->alloc_type; |
| 526 | } |
| 527 | |
| 528 | static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type) |
| 529 | { |
| 530 | struct curseg_info *curseg = CURSEG_I(sbi, type); |
| 531 | return curseg->next_blkoff; |
| 532 | } |
| 533 | |
Jaegeuk Kim | 5d56b67 | 2013-10-29 15:14:54 +0900 | [diff] [blame] | 534 | #ifdef CONFIG_F2FS_CHECK_FS |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 535 | static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) |
| 536 | { |
| 537 | unsigned int end_segno = SM_I(sbi)->segment_count - 1; |
| 538 | BUG_ON(segno > end_segno); |
| 539 | } |
| 540 | |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 541 | static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr) |
| 542 | { |
| 543 | struct f2fs_sm_info *sm_info = SM_I(sbi); |
| 544 | block_t total_blks = sm_info->segment_count << sbi->log_blocks_per_seg; |
| 545 | block_t start_addr = sm_info->seg0_blkaddr; |
| 546 | block_t end_addr = start_addr + total_blks - 1; |
| 547 | BUG_ON(blk_addr < start_addr); |
| 548 | BUG_ON(blk_addr > end_addr); |
| 549 | } |
| 550 | |
| 551 | /* |
| 552 | * Summary block is always treated as invalid block |
| 553 | */ |
| 554 | static inline void check_block_count(struct f2fs_sb_info *sbi, |
| 555 | int segno, struct f2fs_sit_entry *raw_sit) |
| 556 | { |
| 557 | struct f2fs_sm_info *sm_info = SM_I(sbi); |
| 558 | unsigned int end_segno = sm_info->segment_count - 1; |
Chao Yu | 44c60bf | 2013-10-29 14:50:40 +0800 | [diff] [blame] | 559 | bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 560 | int valid_blocks = 0; |
Chao Yu | 44c60bf | 2013-10-29 14:50:40 +0800 | [diff] [blame] | 561 | int cur_pos = 0, next_pos; |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 562 | |
| 563 | /* check segment usage */ |
| 564 | BUG_ON(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg); |
| 565 | |
| 566 | /* check boundary of a given segment number */ |
| 567 | BUG_ON(segno > end_segno); |
| 568 | |
| 569 | /* check bitmap with valid block count */ |
Chao Yu | 44c60bf | 2013-10-29 14:50:40 +0800 | [diff] [blame] | 570 | do { |
| 571 | if (is_valid) { |
| 572 | next_pos = find_next_zero_bit_le(&raw_sit->valid_map, |
| 573 | sbi->blocks_per_seg, |
| 574 | cur_pos); |
| 575 | valid_blocks += next_pos - cur_pos; |
| 576 | } else |
| 577 | next_pos = find_next_bit_le(&raw_sit->valid_map, |
| 578 | sbi->blocks_per_seg, |
| 579 | cur_pos); |
| 580 | cur_pos = next_pos; |
| 581 | is_valid = !is_valid; |
| 582 | } while (cur_pos < sbi->blocks_per_seg); |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 583 | BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks); |
| 584 | } |
Jaegeuk Kim | 5d56b67 | 2013-10-29 15:14:54 +0900 | [diff] [blame] | 585 | #else |
| 586 | #define check_seg_range(sbi, segno) |
| 587 | #define verify_block_addr(sbi, blk_addr) |
| 588 | #define check_block_count(sbi, segno, raw_sit) |
| 589 | #endif |
Jaegeuk Kim | 39a53e0 | 2012-11-28 13:37:31 +0900 | [diff] [blame] | 590 | |
| 591 | static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, |
| 592 | unsigned int start) |
| 593 | { |
| 594 | struct sit_info *sit_i = SIT_I(sbi); |
| 595 | unsigned int offset = SIT_BLOCK_OFFSET(sit_i, start); |
| 596 | block_t blk_addr = sit_i->sit_base_addr + offset; |
| 597 | |
| 598 | check_seg_range(sbi, start); |
| 599 | |
| 600 | /* calculate sit block address */ |
| 601 | if (f2fs_test_bit(offset, sit_i->sit_bitmap)) |
| 602 | blk_addr += sit_i->sit_blocks; |
| 603 | |
| 604 | return blk_addr; |
| 605 | } |
| 606 | |
| 607 | static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, |
| 608 | pgoff_t block_addr) |
| 609 | { |
| 610 | struct sit_info *sit_i = SIT_I(sbi); |
| 611 | block_addr -= sit_i->sit_base_addr; |
| 612 | if (block_addr < sit_i->sit_blocks) |
| 613 | block_addr += sit_i->sit_blocks; |
| 614 | else |
| 615 | block_addr -= sit_i->sit_blocks; |
| 616 | |
| 617 | return block_addr + sit_i->sit_base_addr; |
| 618 | } |
| 619 | |
| 620 | static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) |
| 621 | { |
| 622 | unsigned int block_off = SIT_BLOCK_OFFSET(sit_i, start); |
| 623 | |
| 624 | if (f2fs_test_bit(block_off, sit_i->sit_bitmap)) |
| 625 | f2fs_clear_bit(block_off, sit_i->sit_bitmap); |
| 626 | else |
| 627 | f2fs_set_bit(block_off, sit_i->sit_bitmap); |
| 628 | } |
| 629 | |
| 630 | static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi) |
| 631 | { |
| 632 | struct sit_info *sit_i = SIT_I(sbi); |
| 633 | return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec - |
| 634 | sit_i->mounted_time; |
| 635 | } |
| 636 | |
| 637 | static inline void set_summary(struct f2fs_summary *sum, nid_t nid, |
| 638 | unsigned int ofs_in_node, unsigned char version) |
| 639 | { |
| 640 | sum->nid = cpu_to_le32(nid); |
| 641 | sum->ofs_in_node = cpu_to_le16(ofs_in_node); |
| 642 | sum->version = version; |
| 643 | } |
| 644 | |
| 645 | static inline block_t start_sum_block(struct f2fs_sb_info *sbi) |
| 646 | { |
| 647 | return __start_cp_addr(sbi) + |
| 648 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); |
| 649 | } |
| 650 | |
| 651 | static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) |
| 652 | { |
| 653 | return __start_cp_addr(sbi) + |
| 654 | le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) |
| 655 | - (base + 1) + type; |
| 656 | } |
Jaegeuk Kim | 5ec4e49 | 2013-03-31 13:26:03 +0900 | [diff] [blame] | 657 | |
| 658 | static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) |
| 659 | { |
| 660 | if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) |
| 661 | return true; |
| 662 | return false; |
| 663 | } |
Jaegeuk Kim | ac5d156 | 2013-04-29 16:58:39 +0900 | [diff] [blame] | 664 | |
| 665 | static inline unsigned int max_hw_blocks(struct f2fs_sb_info *sbi) |
| 666 | { |
| 667 | struct block_device *bdev = sbi->sb->s_bdev; |
| 668 | struct request_queue *q = bdev_get_queue(bdev); |
| 669 | return SECTOR_TO_BLOCK(sbi, queue_max_sectors(q)); |
| 670 | } |
Jaegeuk Kim | 87d6f89 | 2014-03-18 12:40:49 +0900 | [diff] [blame] | 671 | |
| 672 | /* |
| 673 | * It is very important to gather dirty pages and write at once, so that we can |
| 674 | * submit a big bio without interfering other data writes. |
| 675 | * By default, 512 pages for directory data, |
| 676 | * 512 pages (2MB) * 3 for three types of nodes, and |
| 677 | * max_bio_blocks for meta are set. |
| 678 | */ |
| 679 | static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) |
| 680 | { |
| 681 | if (type == DATA) |
| 682 | return sbi->blocks_per_seg; |
| 683 | else if (type == NODE) |
| 684 | return 3 * sbi->blocks_per_seg; |
| 685 | else if (type == META) |
| 686 | return MAX_BIO_BLOCKS(max_hw_blocks(sbi)); |
| 687 | else |
| 688 | return 0; |
| 689 | } |
Jaegeuk Kim | 50c8cdb | 2014-03-18 13:47:11 +0900 | [diff] [blame] | 690 | |
| 691 | /* |
| 692 | * When writing pages, it'd better align nr_to_write for segment size. |
| 693 | */ |
| 694 | static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, |
| 695 | struct writeback_control *wbc) |
| 696 | { |
| 697 | long nr_to_write, desired; |
| 698 | |
| 699 | if (wbc->sync_mode != WB_SYNC_NONE) |
| 700 | return 0; |
| 701 | |
| 702 | nr_to_write = wbc->nr_to_write; |
| 703 | |
| 704 | if (type == DATA) |
| 705 | desired = 4096; |
| 706 | else if (type == NODE) |
| 707 | desired = 3 * max_hw_blocks(sbi); |
| 708 | else |
| 709 | desired = MAX_BIO_BLOCKS(max_hw_blocks(sbi)); |
| 710 | |
| 711 | wbc->nr_to_write = desired; |
| 712 | return desired - nr_to_write; |
| 713 | } |