1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * fs/f2fs/segment.h 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/blkdev.h> 9 #include <linux/backing-dev.h> 10 11 /* constant macro */ 12 #define NULL_SEGNO ((unsigned int)(~0)) 13 #define NULL_SECNO ((unsigned int)(~0)) 14 15 #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */ 16 #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */ 17 18 #define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */ 19 #define F2FS_MIN_META_SEGMENTS 8 /* SB + 2 (CP + SIT + NAT) + SSA */ 20 21 /* L: Logical segment # in volume, R: Relative segment # in main area */ 22 #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno) 23 #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno) 24 25 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA) 26 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE) 27 #define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA)) 28 29 static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi, 30 unsigned short seg_type) 31 { 32 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG); 33 } 34 35 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA) 36 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA) 37 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA) 38 39 #define IS_CURSEG(sbi, seg) \ 40 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \ 41 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \ 42 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \ 43 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \ 44 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \ 45 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) || \ 46 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) || \ 47 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno)) 48 49 #define IS_CURSEC(sbi, secno) \ 50 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \ 51 SEGS_PER_SEC(sbi)) || \ 52 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \ 53 SEGS_PER_SEC(sbi)) || \ 54 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \ 55 SEGS_PER_SEC(sbi)) || \ 56 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \ 57 SEGS_PER_SEC(sbi)) || \ 58 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \ 59 SEGS_PER_SEC(sbi)) || \ 60 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \ 61 SEGS_PER_SEC(sbi)) || \ 62 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno / \ 63 SEGS_PER_SEC(sbi)) || \ 64 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno / \ 65 SEGS_PER_SEC(sbi))) 66 67 #define MAIN_BLKADDR(sbi) \ 68 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \ 69 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr)) 70 #define SEG0_BLKADDR(sbi) \ 71 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \ 72 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr)) 73 74 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments) 75 #define MAIN_SECS(sbi) ((sbi)->total_sections) 76 77 #define TOTAL_SEGS(sbi) \ 78 (SM_I(sbi) ? SM_I(sbi)->segment_count : \ 79 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count)) 80 #define TOTAL_BLKS(sbi) (SEGS_TO_BLKS(sbi, TOTAL_SEGS(sbi))) 81 82 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi)) 83 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \ 84 (sbi)->log_blocks_per_seg)) 85 86 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \ 87 (SEGS_TO_BLKS(sbi, GET_R2L_SEGNO(FREE_I(sbi), segno)))) 88 89 #define NEXT_FREE_BLKADDR(sbi, curseg) \ 90 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff) 91 92 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi)) 93 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ 94 (BLKS_TO_SEGS(sbi, GET_SEGOFF_FROM_SEG0(sbi, blk_addr))) 95 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ 96 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (BLKS_PER_SEG(sbi) - 1)) 97 98 #define GET_SEGNO(sbi, blk_addr) \ 99 ((!__is_valid_data_blkaddr(blk_addr)) ? \ 100 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ 101 GET_SEGNO_FROM_SEG0(sbi, blk_addr))) 102 #define CAP_BLKS_PER_SEC(sbi) \ 103 (BLKS_PER_SEC(sbi) - (sbi)->unusable_blocks_per_sec) 104 #define CAP_SEGS_PER_SEC(sbi) \ 105 (SEGS_PER_SEC(sbi) - \ 106 BLKS_TO_SEGS(sbi, (sbi)->unusable_blocks_per_sec)) 107 #define GET_SEC_FROM_SEG(sbi, segno) \ 108 (((segno) == -1) ? -1 : (segno) / SEGS_PER_SEC(sbi)) 109 #define GET_SEG_FROM_SEC(sbi, secno) \ 110 ((secno) * SEGS_PER_SEC(sbi)) 111 #define GET_ZONE_FROM_SEC(sbi, secno) \ 112 (((secno) == -1) ? -1 : (secno) / (sbi)->secs_per_zone) 113 #define GET_ZONE_FROM_SEG(sbi, segno) \ 114 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) 115 116 #define GET_SUM_BLOCK(sbi, segno) \ 117 ((sbi)->sm_info->ssa_blkaddr + (segno)) 118 119 #define GET_SUM_TYPE(footer) ((footer)->entry_type) 120 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type)) 121 122 #define SIT_ENTRY_OFFSET(sit_i, segno) \ 123 ((segno) % (sit_i)->sents_per_block) 124 #define SIT_BLOCK_OFFSET(segno) \ 125 ((segno) / SIT_ENTRY_PER_BLOCK) 126 #define START_SEGNO(segno) \ 127 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK) 128 #define SIT_BLK_CNT(sbi) \ 129 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK) 130 #define f2fs_bitmap_size(nr) \ 131 (BITS_TO_LONGS(nr) * sizeof(unsigned long)) 132 133 #define SECTOR_FROM_BLOCK(blk_addr) \ 134 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK) 135 #define SECTOR_TO_BLOCK(sectors) \ 136 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK) 137 138 /* 139 * In the victim_sel_policy->alloc_mode, there are three block allocation modes. 140 * LFS writes data sequentially with cleaning operations. 141 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. 142 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into 143 * fragmented segment which has similar aging degree. 144 */ 145 enum { 146 LFS = 0, 147 SSR, 148 AT_SSR, 149 }; 150 151 /* 152 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes. 153 * GC_CB is based on cost-benefit algorithm. 154 * GC_GREEDY is based on greedy algorithm. 155 * GC_AT is based on age-threshold algorithm. 156 */ 157 enum { 158 GC_CB = 0, 159 GC_GREEDY, 160 GC_AT, 161 ALLOC_NEXT, 162 FLUSH_DEVICE, 163 MAX_GC_POLICY, 164 }; 165 166 /* 167 * BG_GC means the background cleaning job. 168 * FG_GC means the on-demand cleaning job. 169 */ 170 enum { 171 BG_GC = 0, 172 FG_GC, 173 }; 174 175 /* for a function parameter to select a victim segment */ 176 struct victim_sel_policy { 177 int alloc_mode; /* LFS or SSR */ 178 int gc_mode; /* GC_CB or GC_GREEDY */ 179 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */ 180 unsigned int max_search; /* 181 * maximum # of segments/sections 182 * to search 183 */ 184 unsigned int offset; /* last scanned bitmap offset */ 185 unsigned int ofs_unit; /* bitmap search unit */ 186 unsigned int min_cost; /* minimum cost */ 187 unsigned long long oldest_age; /* oldest age of segments having the same min cost */ 188 unsigned int min_segno; /* segment # having min. cost */ 189 unsigned long long age; /* mtime of GCed section*/ 190 unsigned long long age_threshold;/* age threshold */ 191 bool one_time_gc; /* one time GC */ 192 }; 193 194 struct seg_entry { 195 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ 196 unsigned int valid_blocks:10; /* # of valid blocks */ 197 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ 198 unsigned int padding:6; /* padding */ 199 unsigned char *cur_valid_map; /* validity bitmap of blocks */ 200 #ifdef CONFIG_F2FS_CHECK_FS 201 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ 202 #endif 203 /* 204 * # of valid blocks and the validity bitmap stored in the last 205 * checkpoint pack. This information is used by the SSR mode. 206 */ 207 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ 208 unsigned char *discard_map; 209 unsigned long long mtime; /* modification time of the segment */ 210 }; 211 212 struct sec_entry { 213 unsigned int valid_blocks; /* # of valid blocks in a section */ 214 }; 215 216 #define MAX_SKIP_GC_COUNT 16 217 218 struct revoke_entry { 219 struct list_head list; 220 block_t old_addr; /* for revoking when fail to commit */ 221 pgoff_t index; 222 }; 223 224 struct sit_info { 225 block_t sit_base_addr; /* start block address of SIT area */ 226 block_t sit_blocks; /* # of blocks used by SIT area */ 227 block_t written_valid_blocks; /* # of valid blocks in main area */ 228 char *bitmap; /* all bitmaps pointer */ 229 char *sit_bitmap; /* SIT bitmap pointer */ 230 #ifdef CONFIG_F2FS_CHECK_FS 231 char *sit_bitmap_mir; /* SIT bitmap mirror */ 232 233 /* bitmap of segments to be ignored by GC in case of errors */ 234 unsigned long *invalid_segmap; 235 #endif 236 unsigned int bitmap_size; /* SIT bitmap size */ 237 238 unsigned long *tmp_map; /* bitmap for temporal use */ 239 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ 240 unsigned int dirty_sentries; /* # of dirty sentries */ 241 unsigned int sents_per_block; /* # of SIT entries per block */ 242 struct rw_semaphore sentry_lock; /* to protect SIT cache */ 243 struct seg_entry *sentries; /* SIT segment-level cache */ 244 struct sec_entry *sec_entries; /* SIT section-level cache */ 245 246 /* for cost-benefit algorithm in cleaning procedure */ 247 unsigned long long elapsed_time; /* elapsed time after mount */ 248 unsigned long long mounted_time; /* mount time */ 249 unsigned long long min_mtime; /* min. modification time */ 250 unsigned long long max_mtime; /* max. modification time */ 251 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */ 252 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */ 253 254 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */ 255 }; 256 257 struct free_segmap_info { 258 unsigned int start_segno; /* start segment number logically */ 259 unsigned int free_segments; /* # of free segments */ 260 unsigned int free_sections; /* # of free sections */ 261 spinlock_t segmap_lock; /* free segmap lock */ 262 unsigned long *free_segmap; /* free segment bitmap */ 263 unsigned long *free_secmap; /* free section bitmap */ 264 }; 265 266 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ 267 enum dirty_type { 268 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ 269 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ 270 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ 271 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ 272 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ 273 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ 274 DIRTY, /* to count # of dirty segments */ 275 PRE, /* to count # of entirely obsolete segments */ 276 NR_DIRTY_TYPE 277 }; 278 279 struct dirty_seglist_info { 280 unsigned long *dirty_segmap[NR_DIRTY_TYPE]; 281 unsigned long *dirty_secmap; 282 struct mutex seglist_lock; /* lock for segment bitmaps */ 283 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ 284 unsigned long *victim_secmap; /* background GC victims */ 285 unsigned long *pinned_secmap; /* pinned victims from foreground GC */ 286 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */ 287 bool enable_pin_section; /* enable pinning section */ 288 }; 289 290 /* for active log information */ 291 struct curseg_info { 292 struct mutex curseg_mutex; /* lock for consistency */ 293 struct f2fs_summary_block *sum_blk; /* cached summary block */ 294 struct rw_semaphore journal_rwsem; /* protect journal area */ 295 struct f2fs_journal *journal; /* cached journal info */ 296 unsigned char alloc_type; /* current allocation type */ 297 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */ 298 unsigned int segno; /* current segment number */ 299 unsigned short next_blkoff; /* next block offset to write */ 300 unsigned int zone; /* current zone number */ 301 unsigned int next_segno; /* preallocated segment */ 302 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */ 303 bool inited; /* indicate inmem log is inited */ 304 }; 305 306 struct sit_entry_set { 307 struct list_head set_list; /* link with all sit sets */ 308 unsigned int start_segno; /* start segno of sits in set */ 309 unsigned int entry_cnt; /* the # of sit entries in set */ 310 }; 311 312 /* 313 * inline functions 314 */ 315 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) 316 { 317 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); 318 } 319 320 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, 321 unsigned int segno) 322 { 323 struct sit_info *sit_i = SIT_I(sbi); 324 return &sit_i->sentries[segno]; 325 } 326 327 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, 328 unsigned int segno) 329 { 330 struct sit_info *sit_i = SIT_I(sbi); 331 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; 332 } 333 334 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, 335 unsigned int segno, bool use_section) 336 { 337 /* 338 * In order to get # of valid blocks in a section instantly from many 339 * segments, f2fs manages two counting structures separately. 340 */ 341 if (use_section && __is_large_section(sbi)) 342 return get_sec_entry(sbi, segno)->valid_blocks; 343 else 344 return get_seg_entry(sbi, segno)->valid_blocks; 345 } 346 347 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi, 348 unsigned int segno, bool use_section) 349 { 350 if (use_section && __is_large_section(sbi)) { 351 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 352 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 353 unsigned int blocks = 0; 354 int i; 355 356 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) { 357 struct seg_entry *se = get_seg_entry(sbi, start_segno); 358 359 blocks += se->ckpt_valid_blocks; 360 } 361 return blocks; 362 } 363 return get_seg_entry(sbi, segno)->ckpt_valid_blocks; 364 } 365 366 static inline void seg_info_from_raw_sit(struct seg_entry *se, 367 struct f2fs_sit_entry *rs) 368 { 369 se->valid_blocks = GET_SIT_VBLOCKS(rs); 370 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); 371 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 372 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 373 #ifdef CONFIG_F2FS_CHECK_FS 374 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 375 #endif 376 se->type = GET_SIT_TYPE(rs); 377 se->mtime = le64_to_cpu(rs->mtime); 378 } 379 380 static inline void __seg_info_to_raw_sit(struct seg_entry *se, 381 struct f2fs_sit_entry *rs) 382 { 383 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | 384 se->valid_blocks; 385 rs->vblocks = cpu_to_le16(raw_vblocks); 386 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); 387 rs->mtime = cpu_to_le64(se->mtime); 388 } 389 390 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi, 391 struct page *page, unsigned int start) 392 { 393 struct f2fs_sit_block *raw_sit; 394 struct seg_entry *se; 395 struct f2fs_sit_entry *rs; 396 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK, 397 (unsigned long)MAIN_SEGS(sbi)); 398 int i; 399 400 raw_sit = (struct f2fs_sit_block *)page_address(page); 401 memset(raw_sit, 0, PAGE_SIZE); 402 for (i = 0; i < end - start; i++) { 403 rs = &raw_sit->entries[i]; 404 se = get_seg_entry(sbi, start + i); 405 __seg_info_to_raw_sit(se, rs); 406 } 407 } 408 409 static inline void seg_info_to_raw_sit(struct seg_entry *se, 410 struct f2fs_sit_entry *rs) 411 { 412 __seg_info_to_raw_sit(se, rs); 413 414 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 415 se->ckpt_valid_blocks = se->valid_blocks; 416 } 417 418 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, 419 unsigned int max, unsigned int segno) 420 { 421 unsigned int ret; 422 spin_lock(&free_i->segmap_lock); 423 ret = find_next_bit(free_i->free_segmap, max, segno); 424 spin_unlock(&free_i->segmap_lock); 425 return ret; 426 } 427 428 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) 429 { 430 struct free_segmap_info *free_i = FREE_I(sbi); 431 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 432 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 433 unsigned int next; 434 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi); 435 436 spin_lock(&free_i->segmap_lock); 437 clear_bit(segno, free_i->free_segmap); 438 free_i->free_segments++; 439 440 next = find_next_bit(free_i->free_segmap, 441 start_segno + SEGS_PER_SEC(sbi), start_segno); 442 if (next >= start_segno + usable_segs) { 443 clear_bit(secno, free_i->free_secmap); 444 free_i->free_sections++; 445 } 446 spin_unlock(&free_i->segmap_lock); 447 } 448 449 static inline void __set_inuse(struct f2fs_sb_info *sbi, 450 unsigned int segno) 451 { 452 struct free_segmap_info *free_i = FREE_I(sbi); 453 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 454 455 set_bit(segno, free_i->free_segmap); 456 free_i->free_segments--; 457 if (!test_and_set_bit(secno, free_i->free_secmap)) 458 free_i->free_sections--; 459 } 460 461 static inline void __set_test_and_free(struct f2fs_sb_info *sbi, 462 unsigned int segno, bool inmem) 463 { 464 struct free_segmap_info *free_i = FREE_I(sbi); 465 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 466 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 467 unsigned int next; 468 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi); 469 470 spin_lock(&free_i->segmap_lock); 471 if (test_and_clear_bit(segno, free_i->free_segmap)) { 472 free_i->free_segments++; 473 474 if (!inmem && IS_CURSEC(sbi, secno)) 475 goto skip_free; 476 next = find_next_bit(free_i->free_segmap, 477 start_segno + SEGS_PER_SEC(sbi), start_segno); 478 if (next >= start_segno + usable_segs) { 479 if (test_and_clear_bit(secno, free_i->free_secmap)) 480 free_i->free_sections++; 481 } 482 } 483 skip_free: 484 spin_unlock(&free_i->segmap_lock); 485 } 486 487 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, 488 unsigned int segno) 489 { 490 struct free_segmap_info *free_i = FREE_I(sbi); 491 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 492 493 spin_lock(&free_i->segmap_lock); 494 if (!test_and_set_bit(segno, free_i->free_segmap)) { 495 free_i->free_segments--; 496 if (!test_and_set_bit(secno, free_i->free_secmap)) 497 free_i->free_sections--; 498 } 499 spin_unlock(&free_i->segmap_lock); 500 } 501 502 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, 503 void *dst_addr) 504 { 505 struct sit_info *sit_i = SIT_I(sbi); 506 507 #ifdef CONFIG_F2FS_CHECK_FS 508 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, 509 sit_i->bitmap_size)) 510 f2fs_bug_on(sbi, 1); 511 #endif 512 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); 513 } 514 515 static inline block_t written_block_count(struct f2fs_sb_info *sbi) 516 { 517 return SIT_I(sbi)->written_valid_blocks; 518 } 519 520 static inline unsigned int free_segments(struct f2fs_sb_info *sbi) 521 { 522 return FREE_I(sbi)->free_segments; 523 } 524 525 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi) 526 { 527 return SM_I(sbi)->reserved_segments + 528 SM_I(sbi)->additional_reserved_segments; 529 } 530 531 static inline unsigned int free_sections(struct f2fs_sb_info *sbi) 532 { 533 return FREE_I(sbi)->free_sections; 534 } 535 536 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) 537 { 538 return DIRTY_I(sbi)->nr_dirty[PRE]; 539 } 540 541 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) 542 { 543 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + 544 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + 545 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + 546 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + 547 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + 548 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; 549 } 550 551 static inline int overprovision_segments(struct f2fs_sb_info *sbi) 552 { 553 return SM_I(sbi)->ovp_segments; 554 } 555 556 static inline int reserved_sections(struct f2fs_sb_info *sbi) 557 { 558 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi)); 559 } 560 561 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi, 562 unsigned int node_blocks, unsigned int dent_blocks) 563 { 564 565 unsigned segno, left_blocks; 566 int i; 567 568 /* check current node sections in the worst case. */ 569 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) { 570 segno = CURSEG_I(sbi, i)->segno; 571 left_blocks = CAP_BLKS_PER_SEC(sbi) - 572 get_ckpt_valid_blocks(sbi, segno, true); 573 if (node_blocks > left_blocks) 574 return false; 575 } 576 577 /* check current data section for dentry blocks. */ 578 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno; 579 left_blocks = CAP_BLKS_PER_SEC(sbi) - 580 get_ckpt_valid_blocks(sbi, segno, true); 581 if (dent_blocks > left_blocks) 582 return false; 583 return true; 584 } 585 586 /* 587 * calculate needed sections for dirty node/dentry 588 * and call has_curseg_enough_space 589 */ 590 static inline void __get_secs_required(struct f2fs_sb_info *sbi, 591 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p) 592 { 593 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) + 594 get_pages(sbi, F2FS_DIRTY_DENTS) + 595 get_pages(sbi, F2FS_DIRTY_IMETA); 596 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS); 597 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi); 598 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi); 599 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi); 600 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi); 601 602 if (lower_p) 603 *lower_p = node_secs + dent_secs; 604 if (upper_p) 605 *upper_p = node_secs + dent_secs + 606 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0); 607 if (curseg_p) 608 *curseg_p = has_curseg_enough_space(sbi, 609 node_blocks, dent_blocks); 610 } 611 612 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, 613 int freed, int needed) 614 { 615 unsigned int free_secs, lower_secs, upper_secs; 616 bool curseg_space; 617 618 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 619 return false; 620 621 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space); 622 623 free_secs = free_sections(sbi) + freed; 624 lower_secs += needed + reserved_sections(sbi); 625 upper_secs += needed + reserved_sections(sbi); 626 627 if (free_secs > upper_secs) 628 return false; 629 if (free_secs <= lower_secs) 630 return true; 631 return !curseg_space; 632 } 633 634 static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi, 635 int freed, int needed) 636 { 637 return !has_not_enough_free_secs(sbi, freed, needed); 638 } 639 640 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi) 641 { 642 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 643 return true; 644 if (likely(has_enough_free_secs(sbi, 0, 0))) 645 return true; 646 return false; 647 } 648 649 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) 650 { 651 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; 652 } 653 654 static inline int utilization(struct f2fs_sb_info *sbi) 655 { 656 return div_u64((u64)valid_user_blocks(sbi) * 100, 657 sbi->user_block_count); 658 } 659 660 /* 661 * Sometimes f2fs may be better to drop out-of-place update policy. 662 * And, users can control the policy through sysfs entries. 663 * There are five policies with triggering conditions as follows. 664 * F2FS_IPU_FORCE - all the time, 665 * F2FS_IPU_SSR - if SSR mode is activated, 666 * F2FS_IPU_UTIL - if FS utilization is over threashold, 667 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over 668 * threashold, 669 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash 670 * storages. IPU will be triggered only if the # of dirty 671 * pages over min_fsync_blocks. (=default option) 672 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests. 673 * F2FS_IPU_NOCACHE - disable IPU bio cache. 674 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has 675 * FI_OPU_WRITE flag. 676 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode) 677 */ 678 #define DEF_MIN_IPU_UTIL 70 679 #define DEF_MIN_FSYNC_BLOCKS 8 680 #define DEF_MIN_HOT_BLOCKS 16 681 682 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */ 683 684 #define F2FS_IPU_DISABLE 0 685 686 /* Modification on enum should be synchronized with ipu_mode_names array */ 687 enum { 688 F2FS_IPU_FORCE, 689 F2FS_IPU_SSR, 690 F2FS_IPU_UTIL, 691 F2FS_IPU_SSR_UTIL, 692 F2FS_IPU_FSYNC, 693 F2FS_IPU_ASYNC, 694 F2FS_IPU_NOCACHE, 695 F2FS_IPU_HONOR_OPU_WRITE, 696 F2FS_IPU_MAX, 697 }; 698 699 static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi) 700 { 701 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE; 702 } 703 704 #define F2FS_IPU_POLICY(name) \ 705 static inline bool IS_##name(struct f2fs_sb_info *sbi) \ 706 { \ 707 return SM_I(sbi)->ipu_policy & BIT(name); \ 708 } 709 710 F2FS_IPU_POLICY(F2FS_IPU_FORCE); 711 F2FS_IPU_POLICY(F2FS_IPU_SSR); 712 F2FS_IPU_POLICY(F2FS_IPU_UTIL); 713 F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL); 714 F2FS_IPU_POLICY(F2FS_IPU_FSYNC); 715 F2FS_IPU_POLICY(F2FS_IPU_ASYNC); 716 F2FS_IPU_POLICY(F2FS_IPU_NOCACHE); 717 F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE); 718 719 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, 720 int type) 721 { 722 struct curseg_info *curseg = CURSEG_I(sbi, type); 723 return curseg->segno; 724 } 725 726 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, 727 int type) 728 { 729 struct curseg_info *curseg = CURSEG_I(sbi, type); 730 return curseg->alloc_type; 731 } 732 733 static inline bool valid_main_segno(struct f2fs_sb_info *sbi, 734 unsigned int segno) 735 { 736 return segno <= (MAIN_SEGS(sbi) - 1); 737 } 738 739 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio) 740 { 741 struct f2fs_sb_info *sbi = fio->sbi; 742 743 if (__is_valid_data_blkaddr(fio->old_blkaddr)) 744 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ? 745 META_GENERIC : DATA_GENERIC); 746 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ? 747 META_GENERIC : DATA_GENERIC_ENHANCE); 748 } 749 750 /* 751 * Summary block is always treated as an invalid block 752 */ 753 static inline int check_block_count(struct f2fs_sb_info *sbi, 754 int segno, struct f2fs_sit_entry *raw_sit) 755 { 756 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; 757 int valid_blocks = 0; 758 int cur_pos = 0, next_pos; 759 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno); 760 761 /* check bitmap with valid block count */ 762 do { 763 if (is_valid) { 764 next_pos = find_next_zero_bit_le(&raw_sit->valid_map, 765 usable_blks_per_seg, 766 cur_pos); 767 valid_blocks += next_pos - cur_pos; 768 } else 769 next_pos = find_next_bit_le(&raw_sit->valid_map, 770 usable_blks_per_seg, 771 cur_pos); 772 cur_pos = next_pos; 773 is_valid = !is_valid; 774 } while (cur_pos < usable_blks_per_seg); 775 776 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) { 777 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d", 778 GET_SIT_VBLOCKS(raw_sit), valid_blocks); 779 set_sbi_flag(sbi, SBI_NEED_FSCK); 780 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 781 return -EFSCORRUPTED; 782 } 783 784 if (usable_blks_per_seg < BLKS_PER_SEG(sbi)) 785 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map, 786 BLKS_PER_SEG(sbi), 787 usable_blks_per_seg) != BLKS_PER_SEG(sbi)); 788 789 /* check segment usage, and check boundary of a given segment number */ 790 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg 791 || !valid_main_segno(sbi, segno))) { 792 f2fs_err(sbi, "Wrong valid blocks %d or segno %u", 793 GET_SIT_VBLOCKS(raw_sit), segno); 794 set_sbi_flag(sbi, SBI_NEED_FSCK); 795 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 796 return -EFSCORRUPTED; 797 } 798 return 0; 799 } 800 801 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, 802 unsigned int start) 803 { 804 struct sit_info *sit_i = SIT_I(sbi); 805 unsigned int offset = SIT_BLOCK_OFFSET(start); 806 block_t blk_addr = sit_i->sit_base_addr + offset; 807 808 f2fs_bug_on(sbi, !valid_main_segno(sbi, start)); 809 810 #ifdef CONFIG_F2FS_CHECK_FS 811 if (f2fs_test_bit(offset, sit_i->sit_bitmap) != 812 f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) 813 f2fs_bug_on(sbi, 1); 814 #endif 815 816 /* calculate sit block address */ 817 if (f2fs_test_bit(offset, sit_i->sit_bitmap)) 818 blk_addr += sit_i->sit_blocks; 819 820 return blk_addr; 821 } 822 823 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, 824 pgoff_t block_addr) 825 { 826 struct sit_info *sit_i = SIT_I(sbi); 827 block_addr -= sit_i->sit_base_addr; 828 if (block_addr < sit_i->sit_blocks) 829 block_addr += sit_i->sit_blocks; 830 else 831 block_addr -= sit_i->sit_blocks; 832 833 return block_addr + sit_i->sit_base_addr; 834 } 835 836 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) 837 { 838 unsigned int block_off = SIT_BLOCK_OFFSET(start); 839 840 f2fs_change_bit(block_off, sit_i->sit_bitmap); 841 #ifdef CONFIG_F2FS_CHECK_FS 842 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); 843 #endif 844 } 845 846 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi, 847 bool base_time) 848 { 849 struct sit_info *sit_i = SIT_I(sbi); 850 time64_t diff, now = ktime_get_boottime_seconds(); 851 852 if (now >= sit_i->mounted_time) 853 return sit_i->elapsed_time + now - sit_i->mounted_time; 854 855 /* system time is set to the past */ 856 if (!base_time) { 857 diff = sit_i->mounted_time - now; 858 if (sit_i->elapsed_time >= diff) 859 return sit_i->elapsed_time - diff; 860 return 0; 861 } 862 return sit_i->elapsed_time; 863 } 864 865 static inline void set_summary(struct f2fs_summary *sum, nid_t nid, 866 unsigned int ofs_in_node, unsigned char version) 867 { 868 sum->nid = cpu_to_le32(nid); 869 sum->ofs_in_node = cpu_to_le16(ofs_in_node); 870 sum->version = version; 871 } 872 873 static inline block_t start_sum_block(struct f2fs_sb_info *sbi) 874 { 875 return __start_cp_addr(sbi) + 876 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); 877 } 878 879 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) 880 { 881 return __start_cp_addr(sbi) + 882 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) 883 - (base + 1) + type; 884 } 885 886 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) 887 { 888 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) 889 return true; 890 return false; 891 } 892 893 /* 894 * It is very important to gather dirty pages and write at once, so that we can 895 * submit a big bio without interfering other data writes. 896 * By default, 512 pages for directory data, 897 * 512 pages (2MB) * 8 for nodes, and 898 * 256 pages * 8 for meta are set. 899 */ 900 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) 901 { 902 if (sbi->sb->s_bdi->wb.dirty_exceeded) 903 return 0; 904 905 if (type == DATA) 906 return BLKS_PER_SEG(sbi); 907 else if (type == NODE) 908 return SEGS_TO_BLKS(sbi, 8); 909 else if (type == META) 910 return 8 * BIO_MAX_VECS; 911 else 912 return 0; 913 } 914 915 /* 916 * When writing pages, it'd better align nr_to_write for segment size. 917 */ 918 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, 919 struct writeback_control *wbc) 920 { 921 long nr_to_write, desired; 922 923 if (wbc->sync_mode != WB_SYNC_NONE) 924 return 0; 925 926 nr_to_write = wbc->nr_to_write; 927 desired = BIO_MAX_VECS; 928 if (type == NODE) 929 desired <<= 1; 930 931 wbc->nr_to_write = desired; 932 return desired - nr_to_write; 933 } 934 935 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force) 936 { 937 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info; 938 bool wakeup = false; 939 int i; 940 941 if (force) 942 goto wake_up; 943 944 mutex_lock(&dcc->cmd_lock); 945 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) { 946 if (i + 1 < dcc->discard_granularity) 947 break; 948 if (!list_empty(&dcc->pend_list[i])) { 949 wakeup = true; 950 break; 951 } 952 } 953 mutex_unlock(&dcc->cmd_lock); 954 if (!wakeup || !is_idle(sbi, DISCARD_TIME)) 955 return; 956 wake_up: 957 dcc->discard_wake = true; 958 wake_up_interruptible_all(&dcc->discard_wait_queue); 959 } 960 961 static inline unsigned int first_zoned_segno(struct f2fs_sb_info *sbi) 962 { 963 int devi; 964 965 for (devi = 0; devi < sbi->s_ndevs; devi++) 966 if (bdev_is_zoned(FDEV(devi).bdev)) 967 return GET_SEGNO(sbi, FDEV(devi).start_blk); 968 return 0; 969 } 970