FILE
segment.h
File preview with syntax highlighting.
linux-6.19.14/fs/f2fs/segment.h
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#define INVALID_MTIME ULLONG_MAX /* no valid blocks in a segment/section */ 22 23/* L: Logical segment # in volume, R: Relative segment # in main area */ 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) 26 27#define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA) 28#define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE) 29#define SE_PAGETYPE(se) ((IS_NODESEG((se)->type) ? NODE : DATA)) 30 31static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi, 32 unsigned short seg_type) 33{ 34 f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG); 35} 36 37#define MAIN_BLKADDR(sbi) \ 38 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \ 39 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr)) 40#define SEG0_BLKADDR(sbi) \ 41 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \ 42 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr)) 43 44#define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments) 45#define MAIN_SECS(sbi) ((sbi)->total_sections) 46 47#define TOTAL_SEGS(sbi) \ 48 (SM_I(sbi) ? SM_I(sbi)->segment_count : \ 49 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count)) 50#define TOTAL_BLKS(sbi) (SEGS_TO_BLKS(sbi, TOTAL_SEGS(sbi))) 51 52#define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi)) 53#define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \ 54 (sbi)->log_blocks_per_seg)) 55 56#define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \ 57 (SEGS_TO_BLKS(sbi, GET_R2L_SEGNO(FREE_I(sbi), segno)))) 58 59#define NEXT_FREE_BLKADDR(sbi, curseg) \ 60 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff) 61 62#define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi)) 63#define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ 64 (BLKS_TO_SEGS(sbi, GET_SEGOFF_FROM_SEG0(sbi, blk_addr))) 65#define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ 66 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (BLKS_PER_SEG(sbi) - 1)) 67 68#define GET_SEGNO(sbi, blk_addr) \ 69 ((!__is_valid_data_blkaddr(blk_addr)) ? \ 70 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ 71 GET_SEGNO_FROM_SEG0(sbi, blk_addr))) 72#ifdef CONFIG_BLK_DEV_ZONED 73#define CAP_BLKS_PER_SEC(sbi) \ 74 (BLKS_PER_SEC(sbi) - (sbi)->unusable_blocks_per_sec) 75#define CAP_SEGS_PER_SEC(sbi) \ 76 (SEGS_PER_SEC(sbi) - \ 77 BLKS_TO_SEGS(sbi, (sbi)->unusable_blocks_per_sec)) 78#else 79#define CAP_BLKS_PER_SEC(sbi) BLKS_PER_SEC(sbi) 80#define CAP_SEGS_PER_SEC(sbi) SEGS_PER_SEC(sbi) 81#endif 82#define GET_START_SEG_FROM_SEC(sbi, segno) \ 83 (rounddown(segno, SEGS_PER_SEC(sbi))) 84#define GET_SEC_FROM_SEG(sbi, segno) \ 85 (((segno) == -1) ? -1 : (segno) / SEGS_PER_SEC(sbi)) 86#define GET_SEG_FROM_SEC(sbi, secno) \ 87 ((secno) * SEGS_PER_SEC(sbi)) 88#define GET_ZONE_FROM_SEC(sbi, secno) \ 89 (((secno) == -1) ? -1 : (secno) / (sbi)->secs_per_zone) 90#define GET_ZONE_FROM_SEG(sbi, segno) \ 91 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) 92 93#define GET_SUM_BLOCK(sbi, segno) \ 94 (SM_I(sbi)->ssa_blkaddr + (segno / (sbi)->sums_per_block)) 95#define GET_SUM_BLKOFF(sbi, segno) (segno % (sbi)->sums_per_block) 96#define SUM_BLK_PAGE_ADDR(sbi, folio, segno) \ 97 (folio_address(folio) + GET_SUM_BLKOFF(sbi, segno) * (sbi)->sum_blocksize) 98 99#define GET_SUM_TYPE(footer) ((footer)->entry_type) 100#define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type)) 101 102#define SIT_ENTRY_OFFSET(sit_i, segno) \ 103 ((segno) % (sit_i)->sents_per_block) 104#define SIT_BLOCK_OFFSET(segno) \ 105 ((segno) / SIT_ENTRY_PER_BLOCK) 106#define START_SEGNO(segno) \ 107 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK) 108#define SIT_BLK_CNT(sbi) \ 109 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK) 110#define f2fs_bitmap_size(nr) \ 111 (BITS_TO_LONGS(nr) * sizeof(unsigned long)) 112 113#define SECTOR_FROM_BLOCK(blk_addr) \ 114 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK) 115#define SECTOR_TO_BLOCK(sectors) \ 116 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK) 117 118/* 119 * In the victim_sel_policy->alloc_mode, there are three block allocation modes. 120 * LFS writes data sequentially with cleaning operations. 121 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. 122 * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into 123 * fragmented segment which has similar aging degree. 124 */ 125enum { 126 LFS = 0, 127 SSR, 128 AT_SSR, 129}; 130 131/* 132 * In the victim_sel_policy->gc_mode, there are three gc, aka cleaning, modes. 133 * GC_CB is based on cost-benefit algorithm. 134 * GC_GREEDY is based on greedy algorithm. 135 * GC_AT is based on age-threshold algorithm. 136 */ 137enum { 138 GC_CB = 0, 139 GC_GREEDY, 140 GC_AT, 141 ALLOC_NEXT, 142 FLUSH_DEVICE, 143 MAX_GC_POLICY, 144}; 145 146/* 147 * BG_GC means the background cleaning job. 148 * FG_GC means the on-demand cleaning job. 149 */ 150enum { 151 BG_GC = 0, 152 FG_GC, 153}; 154 155/* for a function parameter to select a victim segment */ 156struct victim_sel_policy { 157 int alloc_mode; /* LFS or SSR */ 158 int gc_mode; /* GC_CB or GC_GREEDY */ 159 unsigned long *dirty_bitmap; /* dirty segment/section bitmap */ 160 unsigned int max_search; /* 161 * maximum # of segments/sections 162 * to search 163 */ 164 unsigned int offset; /* last scanned bitmap offset */ 165 unsigned int ofs_unit; /* bitmap search unit */ 166 unsigned int min_cost; /* minimum cost */ 167 unsigned long long oldest_age; /* oldest age of segments having the same min cost */ 168 unsigned int min_segno; /* segment # having min. cost */ 169 unsigned long long age; /* mtime of GCed section*/ 170 unsigned long long age_threshold;/* age threshold */ 171 bool one_time_gc; /* one time GC */ 172}; 173 174struct seg_entry { 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 */ 179 unsigned char *cur_valid_map; /* validity bitmap of blocks */ 180#ifdef CONFIG_F2FS_CHECK_FS 181 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ 182#endif 183 /* 184 * # of valid blocks and the validity bitmap stored in the last 185 * checkpoint pack. This information is used by the SSR mode. 186 */ 187 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ 188 unsigned char *discard_map; 189 unsigned long long mtime; /* modification time of the segment */ 190}; 191 192struct sec_entry { 193 unsigned int valid_blocks; /* # of valid blocks in a section */ 194 unsigned int ckpt_valid_blocks; /* # of valid blocks last cp in a section */ 195}; 196 197#define MAX_SKIP_GC_COUNT 16 198 199struct revoke_entry { 200 struct list_head list; 201 block_t old_addr; /* for revoking when fail to commit */ 202 pgoff_t index; 203}; 204 205struct sit_info { 206 block_t sit_base_addr; /* start block address of SIT area */ 207 block_t sit_blocks; /* # of blocks used by SIT area */ 208 block_t written_valid_blocks; /* # of valid blocks in main area */ 209 char *bitmap; /* all bitmaps pointer */ 210 char *sit_bitmap; /* SIT bitmap pointer */ 211#ifdef CONFIG_F2FS_CHECK_FS 212 char *sit_bitmap_mir; /* SIT bitmap mirror */ 213 214 /* bitmap of segments to be ignored by GC in case of errors */ 215 unsigned long *invalid_segmap; 216#endif 217 unsigned int bitmap_size; /* SIT bitmap size */ 218 219 unsigned long *tmp_map; /* bitmap for temporal use */ 220 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ 221 unsigned int dirty_sentries; /* # of dirty sentries */ 222 unsigned int sents_per_block; /* # of SIT entries per block */ 223 struct rw_semaphore sentry_lock; /* to protect SIT cache */ 224 struct seg_entry *sentries; /* SIT segment-level cache */ 225 struct sec_entry *sec_entries; /* SIT section-level cache */ 226 227 /* for cost-benefit algorithm in cleaning procedure */ 228 unsigned long long elapsed_time; /* elapsed time after mount */ 229 unsigned long long mounted_time; /* mount time */ 230 unsigned long long min_mtime; /* min. modification time */ 231 unsigned long long max_mtime; /* max. modification time */ 232 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */ 233 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */ 234 235 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */ 236}; 237 238struct free_segmap_info { 239 unsigned int start_segno; /* start segment number logically */ 240 unsigned int free_segments; /* # of free segments */ 241 unsigned int free_sections; /* # of free sections */ 242 spinlock_t segmap_lock; /* free segmap lock */ 243 unsigned long *free_segmap; /* free segment bitmap */ 244 unsigned long *free_secmap; /* free section bitmap */ 245}; 246 247/* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ 248enum dirty_type { 249 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ 250 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ 251 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ 252 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ 253 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ 254 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ 255 DIRTY, /* to count # of dirty segments */ 256 PRE, /* to count # of entirely obsolete segments */ 257 NR_DIRTY_TYPE 258}; 259 260struct dirty_seglist_info { 261 unsigned long *dirty_segmap[NR_DIRTY_TYPE]; 262 unsigned long *dirty_secmap; 263 struct mutex seglist_lock; /* lock for segment bitmaps */ 264 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ 265 unsigned long *victim_secmap; /* background GC victims */ 266 unsigned long *pinned_secmap; /* pinned victims from foreground GC */ 267 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */ 268 bool enable_pin_section; /* enable pinning section */ 269}; 270 271/* for active log information */ 272struct curseg_info { 273 struct mutex curseg_mutex; /* lock for consistency */ 274 struct f2fs_summary_block *sum_blk; /* cached summary block */ 275 struct rw_semaphore journal_rwsem; /* protect journal area */ 276 struct f2fs_journal *journal; /* cached journal info */ 277 unsigned char alloc_type; /* current allocation type */ 278 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */ 279 unsigned int segno; /* current segment number */ 280 unsigned short next_blkoff; /* next block offset to write */ 281 unsigned int zone; /* current zone number */ 282 unsigned int next_segno; /* preallocated segment */ 283 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */ 284 bool inited; /* indicate inmem log is inited */ 285}; 286 287struct sit_entry_set { 288 struct list_head set_list; /* link with all sit sets */ 289 unsigned int start_segno; /* start segno of sits in set */ 290 unsigned int entry_cnt; /* the # of sit entries in set */ 291}; 292 293/* 294 * inline functions 295 */ 296static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) 297{ 298 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); 299} 300 301static inline bool is_curseg(struct f2fs_sb_info *sbi, unsigned int segno) 302{ 303 int i; 304 305 for (i = CURSEG_HOT_DATA; i < NO_CHECK_TYPE; i++) { 306 if (segno == CURSEG_I(sbi, i)->segno) 307 return true; 308 } 309 return false; 310} 311 312static inline bool is_cursec(struct f2fs_sb_info *sbi, unsigned int secno) 313{ 314 int i; 315 316 for (i = CURSEG_HOT_DATA; i < NO_CHECK_TYPE; i++) { 317 if (secno == GET_SEC_FROM_SEG(sbi, CURSEG_I(sbi, i)->segno)) 318 return true; 319 } 320 return false; 321} 322 323static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, 324 unsigned int segno) 325{ 326 struct sit_info *sit_i = SIT_I(sbi); 327 return &sit_i->sentries[segno]; 328} 329 330static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, 331 unsigned int segno) 332{ 333 struct sit_info *sit_i = SIT_I(sbi); 334 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; 335} 336 337static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, 338 unsigned int segno, bool use_section) 339{ 340 /* 341 * In order to get # of valid blocks in a section instantly from many 342 * segments, f2fs manages two counting structures separately. 343 */ 344 if (use_section && __is_large_section(sbi)) 345 return get_sec_entry(sbi, segno)->valid_blocks; 346 else 347 return get_seg_entry(sbi, segno)->valid_blocks; 348} 349 350static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi, 351 unsigned int segno, bool use_section) 352{ 353 if (use_section && __is_large_section(sbi)) 354 return get_sec_entry(sbi, segno)->ckpt_valid_blocks; 355 else 356 return get_seg_entry(sbi, segno)->ckpt_valid_blocks; 357} 358 359static inline void set_ckpt_valid_blocks(struct f2fs_sb_info *sbi, 360 unsigned int segno) 361{ 362 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 363 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 364 unsigned int blocks = 0; 365 int i; 366 367 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) { 368 struct seg_entry *se = get_seg_entry(sbi, start_segno); 369 370 blocks += se->ckpt_valid_blocks; 371 } 372 get_sec_entry(sbi, segno)->ckpt_valid_blocks = blocks; 373} 374 375#ifdef CONFIG_F2FS_CHECK_FS 376static inline void sanity_check_valid_blocks(struct f2fs_sb_info *sbi, 377 unsigned int segno) 378{ 379 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 380 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 381 unsigned int blocks = 0; 382 int i; 383 384 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) { 385 struct seg_entry *se = get_seg_entry(sbi, start_segno); 386 387 blocks += se->ckpt_valid_blocks; 388 } 389 390 if (blocks != get_sec_entry(sbi, segno)->ckpt_valid_blocks) { 391 f2fs_err(sbi, 392 "Inconsistent ckpt valid blocks: " 393 "seg entry(%d) vs sec entry(%d) at secno %d", 394 blocks, get_sec_entry(sbi, segno)->ckpt_valid_blocks, secno); 395 f2fs_bug_on(sbi, 1); 396 } 397} 398#else 399static inline void sanity_check_valid_blocks(struct f2fs_sb_info *sbi, 400 unsigned int segno) 401{ 402} 403#endif 404static inline void seg_info_from_raw_sit(struct seg_entry *se, 405 struct f2fs_sit_entry *rs) 406{ 407 se->valid_blocks = GET_SIT_VBLOCKS(rs); 408 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); 409 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 410 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 411#ifdef CONFIG_F2FS_CHECK_FS 412 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 413#endif 414 se->type = GET_SIT_TYPE(rs); 415 se->mtime = le64_to_cpu(rs->mtime); 416} 417 418static inline void __seg_info_to_raw_sit(struct seg_entry *se, 419 struct f2fs_sit_entry *rs) 420{ 421 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | 422 se->valid_blocks; 423 rs->vblocks = cpu_to_le16(raw_vblocks); 424 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); 425 rs->mtime = cpu_to_le64(se->mtime); 426} 427 428static inline void seg_info_to_sit_folio(struct f2fs_sb_info *sbi, 429 struct folio *folio, unsigned int start) 430{ 431 struct f2fs_sit_block *raw_sit; 432 struct seg_entry *se; 433 struct f2fs_sit_entry *rs; 434 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK, 435 (unsigned long)MAIN_SEGS(sbi)); 436 int i; 437 438 raw_sit = folio_address(folio); 439 memset(raw_sit, 0, PAGE_SIZE); 440 for (i = 0; i < end - start; i++) { 441 rs = &raw_sit->entries[i]; 442 se = get_seg_entry(sbi, start + i); 443 __seg_info_to_raw_sit(se, rs); 444 } 445} 446 447static inline void seg_info_to_raw_sit(struct seg_entry *se, 448 struct f2fs_sit_entry *rs) 449{ 450 __seg_info_to_raw_sit(se, rs); 451 452 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 453 se->ckpt_valid_blocks = se->valid_blocks; 454} 455 456static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, 457 unsigned int max, unsigned int segno) 458{ 459 unsigned int ret; 460 spin_lock(&free_i->segmap_lock); 461 ret = find_next_bit(free_i->free_segmap, max, segno); 462 spin_unlock(&free_i->segmap_lock); 463 return ret; 464} 465 466static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) 467{ 468 struct free_segmap_info *free_i = FREE_I(sbi); 469 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 470 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 471 unsigned int next; 472 473 spin_lock(&free_i->segmap_lock); 474 clear_bit(segno, free_i->free_segmap); 475 free_i->free_segments++; 476 477 next = find_next_bit(free_i->free_segmap, 478 start_segno + SEGS_PER_SEC(sbi), start_segno); 479 if (next >= start_segno + f2fs_usable_segs_in_sec(sbi)) { 480 clear_bit(secno, free_i->free_secmap); 481 free_i->free_sections++; 482 } 483 spin_unlock(&free_i->segmap_lock); 484} 485 486static inline void __set_inuse(struct f2fs_sb_info *sbi, 487 unsigned int segno) 488{ 489 struct free_segmap_info *free_i = FREE_I(sbi); 490 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 491 492 set_bit(segno, free_i->free_segmap); 493 free_i->free_segments--; 494 if (!test_and_set_bit(secno, free_i->free_secmap)) 495 free_i->free_sections--; 496} 497 498static inline void __set_test_and_free(struct f2fs_sb_info *sbi, 499 unsigned int segno, bool inmem) 500{ 501 struct free_segmap_info *free_i = FREE_I(sbi); 502 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 503 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 504 unsigned int next; 505 bool ret; 506 507 spin_lock(&free_i->segmap_lock); 508 ret = test_and_clear_bit(segno, free_i->free_segmap); 509 if (!ret) 510 goto unlock_out; 511 512 free_i->free_segments++; 513 514 if (!inmem && is_cursec(sbi, secno)) 515 goto unlock_out; 516 517 /* check large section */ 518 next = find_next_bit(free_i->free_segmap, 519 start_segno + SEGS_PER_SEC(sbi), start_segno); 520 if (next < start_segno + f2fs_usable_segs_in_sec(sbi)) 521 goto unlock_out; 522 523 ret = test_and_clear_bit(secno, free_i->free_secmap); 524 if (!ret) 525 goto unlock_out; 526 527 free_i->free_sections++; 528 529 if (GET_SEC_FROM_SEG(sbi, sbi->next_victim_seg[BG_GC]) == secno) 530 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 531 if (GET_SEC_FROM_SEG(sbi, sbi->next_victim_seg[FG_GC]) == secno) 532 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 533 534unlock_out: 535 spin_unlock(&free_i->segmap_lock); 536} 537 538static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, 539 unsigned int segno) 540{ 541 struct free_segmap_info *free_i = FREE_I(sbi); 542 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 543 544 spin_lock(&free_i->segmap_lock); 545 if (!test_and_set_bit(segno, free_i->free_segmap)) { 546 free_i->free_segments--; 547 if (!test_and_set_bit(secno, free_i->free_secmap)) 548 free_i->free_sections--; 549 } 550 spin_unlock(&free_i->segmap_lock); 551} 552 553static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, 554 void *dst_addr) 555{ 556 struct sit_info *sit_i = SIT_I(sbi); 557 558#ifdef CONFIG_F2FS_CHECK_FS 559 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, 560 sit_i->bitmap_size)) 561 f2fs_bug_on(sbi, 1); 562#endif 563 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); 564} 565 566static inline block_t written_block_count(struct f2fs_sb_info *sbi) 567{ 568 return SIT_I(sbi)->written_valid_blocks; 569} 570 571static inline unsigned int free_segments(struct f2fs_sb_info *sbi) 572{ 573 return FREE_I(sbi)->free_segments; 574} 575 576static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi) 577{ 578 return SM_I(sbi)->reserved_segments; 579} 580 581static inline unsigned int free_sections(struct f2fs_sb_info *sbi) 582{ 583 return FREE_I(sbi)->free_sections; 584} 585 586static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) 587{ 588 return DIRTY_I(sbi)->nr_dirty[PRE]; 589} 590 591static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) 592{ 593 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + 594 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + 595 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + 596 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + 597 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + 598 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; 599} 600 601static inline int overprovision_segments(struct f2fs_sb_info *sbi) 602{ 603 return SM_I(sbi)->ovp_segments; 604} 605 606static inline int reserved_sections(struct f2fs_sb_info *sbi) 607{ 608 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi)); 609} 610 611static inline unsigned int get_left_section_blocks(struct f2fs_sb_info *sbi, 612 enum log_type type, unsigned int segno) 613{ 614 if (f2fs_lfs_mode(sbi)) { 615 unsigned int used_blocks = __is_large_section(sbi) ? SEGS_TO_BLKS(sbi, 616 (segno - GET_START_SEG_FROM_SEC(sbi, segno))) : 0; 617 return CAP_BLKS_PER_SEC(sbi) - used_blocks - 618 CURSEG_I(sbi, type)->next_blkoff; 619 } 620 return CAP_BLKS_PER_SEC(sbi) - get_ckpt_valid_blocks(sbi, segno, true); 621} 622 623static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi, 624 unsigned int node_blocks, unsigned int data_blocks, 625 unsigned int dent_blocks) 626{ 627 unsigned int segno, left_blocks, blocks; 628 int i; 629 630 /* check current data/node sections in the worst case. */ 631 for (i = CURSEG_HOT_DATA; i < NR_PERSISTENT_LOG; i++) { 632 segno = CURSEG_I(sbi, i)->segno; 633 634 if (unlikely(segno == NULL_SEGNO)) 635 return false; 636 637 left_blocks = get_left_section_blocks(sbi, i, segno); 638 639 blocks = i <= CURSEG_COLD_DATA ? data_blocks : node_blocks; 640 if (blocks > left_blocks) 641 return false; 642 } 643 644 /* check current data section for dentry blocks. */ 645 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno; 646 647 if (unlikely(segno == NULL_SEGNO)) 648 return false; 649 650 left_blocks = get_left_section_blocks(sbi, CURSEG_HOT_DATA, segno); 651 652 if (dent_blocks > left_blocks) 653 return false; 654 return true; 655} 656 657/* 658 * calculate needed sections for dirty node/dentry and call 659 * has_curseg_enough_space, please note that, it needs to account 660 * dirty data as well in lfs mode when checkpoint is disabled. 661 */ 662static inline void __get_secs_required(struct f2fs_sb_info *sbi, 663 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p) 664{ 665 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) + 666 get_pages(sbi, F2FS_DIRTY_DENTS) + 667 get_pages(sbi, F2FS_DIRTY_IMETA); 668 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS); 669 unsigned int total_data_blocks = 0; 670 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi); 671 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi); 672 unsigned int data_secs = 0; 673 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi); 674 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi); 675 unsigned int data_blocks = 0; 676 677 if (f2fs_lfs_mode(sbi)) { 678 total_data_blocks = get_pages(sbi, F2FS_DIRTY_DATA); 679 data_secs = total_data_blocks / CAP_BLKS_PER_SEC(sbi); 680 data_blocks = total_data_blocks % CAP_BLKS_PER_SEC(sbi); 681 } 682 683 if (lower_p) 684 *lower_p = node_secs + dent_secs + data_secs; 685 if (upper_p) 686 *upper_p = node_secs + dent_secs + data_secs + 687 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0) + 688 (data_blocks ? 1 : 0); 689 if (curseg_p) 690 *curseg_p = has_curseg_enough_space(sbi, 691 node_blocks, data_blocks, dent_blocks); 692} 693 694static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, 695 int freed, int needed) 696{ 697 unsigned int free_secs, lower_secs, upper_secs; 698 bool curseg_space; 699 700 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 701 return false; 702 703 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space); 704 705 free_secs = free_sections(sbi) + freed; 706 lower_secs += needed + reserved_sections(sbi); 707 upper_secs += needed + reserved_sections(sbi); 708 709 if (free_secs > upper_secs) 710 return false; 711 if (free_secs <= lower_secs) 712 return true; 713 return !curseg_space; 714} 715 716static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi, 717 int freed, int needed) 718{ 719 return !has_not_enough_free_secs(sbi, freed, needed); 720} 721 722static inline bool has_enough_free_blks(struct f2fs_sb_info *sbi) 723{ 724 unsigned int total_free_blocks = 0; 725 unsigned int avail_user_block_count; 726 727 spin_lock(&sbi->stat_lock); 728 729 avail_user_block_count = get_available_block_count(sbi, NULL, true); 730 total_free_blocks = avail_user_block_count - (unsigned int)valid_user_blocks(sbi); 731 732 spin_unlock(&sbi->stat_lock); 733 734 return total_free_blocks > 0; 735} 736 737static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi) 738{ 739 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 740 return true; 741 if (likely(has_enough_free_secs(sbi, 0, 0))) 742 return true; 743 if (!f2fs_lfs_mode(sbi) && 744 likely(has_enough_free_blks(sbi))) 745 return true; 746 return false; 747} 748 749static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) 750{ 751 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; 752} 753 754static inline int utilization(struct f2fs_sb_info *sbi) 755{ 756 return div_u64((u64)valid_user_blocks(sbi) * 100, 757 sbi->user_block_count); 758} 759 760/* 761 * Sometimes f2fs may be better to drop out-of-place update policy. 762 * And, users can control the policy through sysfs entries. 763 * There are five policies with triggering conditions as follows. 764 * F2FS_IPU_FORCE - all the time, 765 * F2FS_IPU_SSR - if SSR mode is activated, 766 * F2FS_IPU_UTIL - if FS utilization is over threashold, 767 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over 768 * threashold, 769 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash 770 * storages. IPU will be triggered only if the # of dirty 771 * pages over min_fsync_blocks. (=default option) 772 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests. 773 * F2FS_IPU_NOCACHE - disable IPU bio cache. 774 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has 775 * FI_OPU_WRITE flag. 776 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode) 777 */ 778#define DEF_MIN_IPU_UTIL 70 779#define DEF_MIN_FSYNC_BLOCKS 8 780#define DEF_MIN_HOT_BLOCKS 16 781 782#define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */ 783 784#define F2FS_IPU_DISABLE 0 785 786/* Modification on enum should be synchronized with ipu_mode_names array */ 787enum { 788 F2FS_IPU_FORCE, 789 F2FS_IPU_SSR, 790 F2FS_IPU_UTIL, 791 F2FS_IPU_SSR_UTIL, 792 F2FS_IPU_FSYNC, 793 F2FS_IPU_ASYNC, 794 F2FS_IPU_NOCACHE, 795 F2FS_IPU_HONOR_OPU_WRITE, 796 F2FS_IPU_MAX, 797}; 798 799static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi) 800{ 801 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE; 802} 803 804#define F2FS_IPU_POLICY(name) \ 805static inline bool IS_##name(struct f2fs_sb_info *sbi) \ 806{ \ 807 return SM_I(sbi)->ipu_policy & BIT(name); \ 808} 809 810F2FS_IPU_POLICY(F2FS_IPU_FORCE); 811F2FS_IPU_POLICY(F2FS_IPU_SSR); 812F2FS_IPU_POLICY(F2FS_IPU_UTIL); 813F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL); 814F2FS_IPU_POLICY(F2FS_IPU_FSYNC); 815F2FS_IPU_POLICY(F2FS_IPU_ASYNC); 816F2FS_IPU_POLICY(F2FS_IPU_NOCACHE); 817F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE); 818 819static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, 820 int type) 821{ 822 struct curseg_info *curseg = CURSEG_I(sbi, type); 823 return curseg->segno; 824} 825 826static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, 827 int type) 828{ 829 struct curseg_info *curseg = CURSEG_I(sbi, type); 830 return curseg->alloc_type; 831} 832 833static inline bool valid_main_segno(struct f2fs_sb_info *sbi, 834 unsigned int segno) 835{ 836 return segno <= (MAIN_SEGS(sbi) - 1); 837} 838 839static inline void verify_fio_blkaddr(struct f2fs_io_info *fio) 840{ 841 struct f2fs_sb_info *sbi = fio->sbi; 842 843 if (__is_valid_data_blkaddr(fio->old_blkaddr)) 844 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ? 845 META_GENERIC : DATA_GENERIC); 846 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ? 847 META_GENERIC : DATA_GENERIC_ENHANCE); 848} 849 850/* 851 * Summary block is always treated as an invalid block 852 */ 853static inline int check_block_count(struct f2fs_sb_info *sbi, 854 int segno, struct f2fs_sit_entry *raw_sit) 855{ 856 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; 857 int valid_blocks = 0; 858 int cur_pos = 0, next_pos; 859 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno); 860 861 /* check bitmap with valid block count */ 862 do { 863 if (is_valid) { 864 next_pos = find_next_zero_bit_le(&raw_sit->valid_map, 865 usable_blks_per_seg, 866 cur_pos); 867 valid_blocks += next_pos - cur_pos; 868 } else 869 next_pos = find_next_bit_le(&raw_sit->valid_map, 870 usable_blks_per_seg, 871 cur_pos); 872 cur_pos = next_pos; 873 is_valid = !is_valid; 874 } while (cur_pos < usable_blks_per_seg); 875 876 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) { 877 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d", 878 GET_SIT_VBLOCKS(raw_sit), valid_blocks); 879 set_sbi_flag(sbi, SBI_NEED_FSCK); 880 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 881 return -EFSCORRUPTED; 882 } 883 884 if (usable_blks_per_seg < BLKS_PER_SEG(sbi)) 885 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map, 886 BLKS_PER_SEG(sbi), 887 usable_blks_per_seg) != BLKS_PER_SEG(sbi)); 888 889 /* check segment usage, and check boundary of a given segment number */ 890 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg 891 || !valid_main_segno(sbi, segno))) { 892 f2fs_err(sbi, "Wrong valid blocks %d or segno %u", 893 GET_SIT_VBLOCKS(raw_sit), segno); 894 set_sbi_flag(sbi, SBI_NEED_FSCK); 895 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 896 return -EFSCORRUPTED; 897 } 898 return 0; 899} 900 901static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, 902 unsigned int start) 903{ 904 struct sit_info *sit_i = SIT_I(sbi); 905 unsigned int offset = SIT_BLOCK_OFFSET(start); 906 block_t blk_addr = sit_i->sit_base_addr + offset; 907 908 f2fs_bug_on(sbi, !valid_main_segno(sbi, start)); 909 910#ifdef CONFIG_F2FS_CHECK_FS 911 if (f2fs_test_bit(offset, sit_i->sit_bitmap) != 912 f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) 913 f2fs_bug_on(sbi, 1); 914#endif 915 916 /* calculate sit block address */ 917 if (f2fs_test_bit(offset, sit_i->sit_bitmap)) 918 blk_addr += sit_i->sit_blocks; 919 920 return blk_addr; 921} 922 923static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, 924 pgoff_t block_addr) 925{ 926 struct sit_info *sit_i = SIT_I(sbi); 927 block_addr -= sit_i->sit_base_addr; 928 if (block_addr < sit_i->sit_blocks) 929 block_addr += sit_i->sit_blocks; 930 else 931 block_addr -= sit_i->sit_blocks; 932 933 return block_addr + sit_i->sit_base_addr; 934} 935 936static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) 937{ 938 unsigned int block_off = SIT_BLOCK_OFFSET(start); 939 940 f2fs_change_bit(block_off, sit_i->sit_bitmap); 941#ifdef CONFIG_F2FS_CHECK_FS 942 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); 943#endif 944} 945 946static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi, 947 bool base_time) 948{ 949 struct sit_info *sit_i = SIT_I(sbi); 950 time64_t diff, now = ktime_get_boottime_seconds(); 951 952 if (now >= sit_i->mounted_time) 953 return sit_i->elapsed_time + now - sit_i->mounted_time; 954 955 /* system time is set to the past */ 956 if (!base_time) { 957 diff = sit_i->mounted_time - now; 958 if (sit_i->elapsed_time >= diff) 959 return sit_i->elapsed_time - diff; 960 return 0; 961 } 962 return sit_i->elapsed_time; 963} 964 965static inline void set_summary(struct f2fs_summary *sum, nid_t nid, 966 unsigned int ofs_in_node, unsigned char version) 967{ 968 sum->nid = cpu_to_le32(nid); 969 sum->ofs_in_node = cpu_to_le16(ofs_in_node); 970 sum->version = version; 971} 972 973static inline block_t start_sum_block(struct f2fs_sb_info *sbi) 974{ 975 return __start_cp_addr(sbi) + 976 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); 977} 978 979static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) 980{ 981 return __start_cp_addr(sbi) + 982 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) 983 - (base + 1) + type; 984} 985 986static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) 987{ 988 if (is_cursec(sbi, secno) || (sbi->cur_victim_sec == secno)) 989 return true; 990 return false; 991} 992 993/* 994 * It is very important to gather dirty pages and write at once, so that we can 995 * submit a big bio without interfering other data writes. 996 * By default, 512 pages for directory data, 997 * 512 pages (2MB) * 8 for nodes, and 998 * 256 pages * 8 for meta are set. 999 */ 1000static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) 1001{ 1002 if (sbi->sb->s_bdi->wb.dirty_exceeded) 1003 return 0; 1004 1005 if (type == DATA) 1006 return BLKS_PER_SEG(sbi); 1007 else if (type == NODE) 1008 return SEGS_TO_BLKS(sbi, 8); 1009 else if (type == META) 1010 return 8 * BIO_MAX_VECS; 1011 else 1012 return 0; 1013} 1014 1015/* 1016 * When writing pages, it'd better align nr_to_write for segment size. 1017 */ 1018static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, 1019 struct writeback_control *wbc) 1020{ 1021 long nr_to_write, desired; 1022 1023 if (wbc->sync_mode != WB_SYNC_NONE) 1024 return 0; 1025 1026 nr_to_write = wbc->nr_to_write; 1027 desired = BIO_MAX_VECS; 1028 if (type == NODE) 1029 desired <<= 1; 1030 1031 wbc->nr_to_write = desired; 1032 return desired - nr_to_write; 1033} 1034 1035static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force) 1036{ 1037 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info; 1038 bool wakeup = false; 1039 int i; 1040 1041 if (force) 1042 goto wake_up; 1043 1044 mutex_lock(&dcc->cmd_lock); 1045 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) { 1046 if (i + 1 < dcc->discard_granularity) 1047 break; 1048 if (!list_empty(&dcc->pend_list[i])) { 1049 wakeup = true; 1050 break; 1051 } 1052 } 1053 mutex_unlock(&dcc->cmd_lock); 1054 if (!wakeup || !is_idle(sbi, DISCARD_TIME)) 1055 return; 1056wake_up: 1057 dcc->discard_wake = true; 1058 wake_up_interruptible_all(&dcc->discard_wait_queue); 1059}
~/
Read-only file preview.