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 20 /* L: Logical segment # in volume, R: Relative segment # in main area */ 21 #define GET_L2R_SEGNO(free_i, segno) ((segno) - (free_i)->start_segno) 22 #define GET_R2L_SEGNO(free_i, segno) ((segno) + (free_i)->start_segno) 23 24 #define IS_DATASEG(t) ((t) <= CURSEG_COLD_DATA) 25 #define IS_NODESEG(t) ((t) >= CURSEG_HOT_NODE) 26 27 #define IS_HOT(t) ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA) 28 #define IS_WARM(t) ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA) 29 #define IS_COLD(t) ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA) 30 31 #define IS_CURSEG(sbi, seg) \ 32 (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \ 33 ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \ 34 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \ 35 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \ 36 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \ 37 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno)) 38 39 #define IS_CURSEC(sbi, secno) \ 40 (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \ 41 (sbi)->segs_per_sec) || \ 42 ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \ 43 (sbi)->segs_per_sec) || \ 44 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \ 45 (sbi)->segs_per_sec) || \ 46 ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \ 47 (sbi)->segs_per_sec) || \ 48 ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \ 49 (sbi)->segs_per_sec) || \ 50 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \ 51 (sbi)->segs_per_sec)) \ 52 53 #define MAIN_BLKADDR(sbi) \ 54 (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \ 55 le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr)) 56 #define SEG0_BLKADDR(sbi) \ 57 (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \ 58 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr)) 59 60 #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments) 61 #define MAIN_SECS(sbi) ((sbi)->total_sections) 62 63 #define TOTAL_SEGS(sbi) \ 64 (SM_I(sbi) ? SM_I(sbi)->segment_count : \ 65 le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count)) 66 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg) 67 68 #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi)) 69 #define SEGMENT_SIZE(sbi) (1ULL << ((sbi)->log_blocksize + \ 70 (sbi)->log_blocks_per_seg)) 71 72 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \ 73 (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg)) 74 75 #define NEXT_FREE_BLKADDR(sbi, curseg) \ 76 (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff) 77 78 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi)) 79 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \ 80 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg) 81 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \ 82 (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1)) 83 84 #define GET_SEGNO(sbi, blk_addr) \ 85 ((!__is_valid_data_blkaddr(blk_addr)) ? \ 86 NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \ 87 GET_SEGNO_FROM_SEG0(sbi, blk_addr))) 88 #define BLKS_PER_SEC(sbi) \ 89 ((sbi)->segs_per_sec * (sbi)->blocks_per_seg) 90 #define GET_SEC_FROM_SEG(sbi, segno) \ 91 ((segno) / (sbi)->segs_per_sec) 92 #define GET_SEG_FROM_SEC(sbi, secno) \ 93 ((secno) * (sbi)->segs_per_sec) 94 #define GET_ZONE_FROM_SEC(sbi, secno) \ 95 ((secno) / (sbi)->secs_per_zone) 96 #define GET_ZONE_FROM_SEG(sbi, segno) \ 97 GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno)) 98 99 #define GET_SUM_BLOCK(sbi, segno) \ 100 ((sbi)->sm_info->ssa_blkaddr + (segno)) 101 102 #define GET_SUM_TYPE(footer) ((footer)->entry_type) 103 #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = (type)) 104 105 #define SIT_ENTRY_OFFSET(sit_i, segno) \ 106 ((segno) % (sit_i)->sents_per_block) 107 #define SIT_BLOCK_OFFSET(segno) \ 108 ((segno) / SIT_ENTRY_PER_BLOCK) 109 #define START_SEGNO(segno) \ 110 (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK) 111 #define SIT_BLK_CNT(sbi) \ 112 DIV_ROUND_UP(MAIN_SEGS(sbi), SIT_ENTRY_PER_BLOCK) 113 #define f2fs_bitmap_size(nr) \ 114 (BITS_TO_LONGS(nr) * sizeof(unsigned long)) 115 116 #define SECTOR_FROM_BLOCK(blk_addr) \ 117 (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK) 118 #define SECTOR_TO_BLOCK(sectors) \ 119 ((sectors) >> F2FS_LOG_SECTORS_PER_BLOCK) 120 121 /* 122 * indicate a block allocation direction: RIGHT and LEFT. 123 * RIGHT means allocating new sections towards the end of volume. 124 * LEFT means the opposite direction. 125 */ 126 enum { 127 ALLOC_RIGHT = 0, 128 ALLOC_LEFT 129 }; 130 131 /* 132 * In the victim_sel_policy->alloc_mode, there are two block allocation modes. 133 * LFS writes data sequentially with cleaning operations. 134 * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations. 135 */ 136 enum { 137 LFS = 0, 138 SSR 139 }; 140 141 /* 142 * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes. 143 * GC_CB is based on cost-benefit algorithm. 144 * GC_GREEDY is based on greedy algorithm. 145 */ 146 enum { 147 GC_CB = 0, 148 GC_GREEDY, 149 ALLOC_NEXT, 150 FLUSH_DEVICE, 151 MAX_GC_POLICY, 152 }; 153 154 /* 155 * BG_GC means the background cleaning job. 156 * FG_GC means the on-demand cleaning job. 157 * FORCE_FG_GC means on-demand cleaning job in background. 158 */ 159 enum { 160 BG_GC = 0, 161 FG_GC, 162 FORCE_FG_GC, 163 }; 164 165 /* for a function parameter to select a victim segment */ 166 struct victim_sel_policy { 167 int alloc_mode; /* LFS or SSR */ 168 int gc_mode; /* GC_CB or GC_GREEDY */ 169 unsigned long *dirty_segmap; /* dirty segment bitmap */ 170 unsigned int max_search; /* maximum # of segments to search */ 171 unsigned int offset; /* last scanned bitmap offset */ 172 unsigned int ofs_unit; /* bitmap search unit */ 173 unsigned int min_cost; /* minimum cost */ 174 unsigned int min_segno; /* segment # having min. cost */ 175 }; 176 177 struct seg_entry { 178 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ 179 unsigned int valid_blocks:10; /* # of valid blocks */ 180 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ 181 unsigned int padding:6; /* padding */ 182 unsigned char *cur_valid_map; /* validity bitmap of blocks */ 183 #ifdef CONFIG_F2FS_CHECK_FS 184 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ 185 #endif 186 /* 187 * # of valid blocks and the validity bitmap stored in the the last 188 * checkpoint pack. This information is used by the SSR mode. 189 */ 190 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ 191 unsigned char *discard_map; 192 unsigned long long mtime; /* modification time of the segment */ 193 }; 194 195 struct sec_entry { 196 unsigned int valid_blocks; /* # of valid blocks in a section */ 197 }; 198 199 struct segment_allocation { 200 void (*allocate_segment)(struct f2fs_sb_info *, int, bool); 201 }; 202 203 /* 204 * this value is set in page as a private data which indicate that 205 * the page is atomically written, and it is in inmem_pages list. 206 */ 207 #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1) 208 #define DUMMY_WRITTEN_PAGE ((unsigned long)-2) 209 210 #define IS_ATOMIC_WRITTEN_PAGE(page) \ 211 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE) 212 #define IS_DUMMY_WRITTEN_PAGE(page) \ 213 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE) 214 215 #define MAX_SKIP_GC_COUNT 16 216 217 struct inmem_pages { 218 struct list_head list; 219 struct page *page; 220 block_t old_addr; /* for revoking when fail to commit */ 221 }; 222 223 struct sit_info { 224 const struct segment_allocation *s_ops; 225 226 block_t sit_base_addr; /* start block address of SIT area */ 227 block_t sit_blocks; /* # of blocks used by SIT area */ 228 block_t written_valid_blocks; /* # of valid blocks in main area */ 229 char *bitmap; /* all bitmaps pointer */ 230 char *sit_bitmap; /* SIT bitmap pointer */ 231 #ifdef CONFIG_F2FS_CHECK_FS 232 char *sit_bitmap_mir; /* SIT bitmap mirror */ 233 234 /* bitmap of segments to be ignored by GC in case of errors */ 235 unsigned long *invalid_segmap; 236 #endif 237 unsigned int bitmap_size; /* SIT bitmap size */ 238 239 unsigned long *tmp_map; /* bitmap for temporal use */ 240 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ 241 unsigned int dirty_sentries; /* # of dirty sentries */ 242 unsigned int sents_per_block; /* # of SIT entries per block */ 243 struct rw_semaphore sentry_lock; /* to protect SIT cache */ 244 struct seg_entry *sentries; /* SIT segment-level cache */ 245 struct sec_entry *sec_entries; /* SIT section-level cache */ 246 247 /* for cost-benefit algorithm in cleaning procedure */ 248 unsigned long long elapsed_time; /* elapsed time after mount */ 249 unsigned long long mounted_time; /* mount time */ 250 unsigned long long min_mtime; /* min. modification time */ 251 unsigned long long max_mtime; /* max. modification time */ 252 253 unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */ 254 }; 255 256 struct free_segmap_info { 257 unsigned int start_segno; /* start segment number logically */ 258 unsigned int free_segments; /* # of free segments */ 259 unsigned int free_sections; /* # of free sections */ 260 spinlock_t segmap_lock; /* free segmap lock */ 261 unsigned long *free_segmap; /* free segment bitmap */ 262 unsigned long *free_secmap; /* free section bitmap */ 263 }; 264 265 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */ 266 enum dirty_type { 267 DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */ 268 DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */ 269 DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */ 270 DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */ 271 DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */ 272 DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */ 273 DIRTY, /* to count # of dirty segments */ 274 PRE, /* to count # of entirely obsolete segments */ 275 NR_DIRTY_TYPE 276 }; 277 278 struct dirty_seglist_info { 279 const struct victim_selection *v_ops; /* victim selction operation */ 280 unsigned long *dirty_segmap[NR_DIRTY_TYPE]; 281 struct mutex seglist_lock; /* lock for segment bitmaps */ 282 int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */ 283 unsigned long *victim_secmap; /* background GC victims */ 284 }; 285 286 /* victim selection function for cleaning and SSR */ 287 struct victim_selection { 288 int (*get_victim)(struct f2fs_sb_info *, unsigned int *, 289 int, int, char); 290 }; 291 292 /* for active log information */ 293 struct curseg_info { 294 struct mutex curseg_mutex; /* lock for consistency */ 295 struct f2fs_summary_block *sum_blk; /* cached summary block */ 296 struct rw_semaphore journal_rwsem; /* protect journal area */ 297 struct f2fs_journal *journal; /* cached journal info */ 298 unsigned char alloc_type; /* current allocation type */ 299 unsigned int segno; /* current segment number */ 300 unsigned short next_blkoff; /* next block offset to write */ 301 unsigned int zone; /* current zone number */ 302 unsigned int next_segno; /* preallocated segment */ 303 }; 304 305 struct sit_entry_set { 306 struct list_head set_list; /* link with all sit sets */ 307 unsigned int start_segno; /* start segno of sits in set */ 308 unsigned int entry_cnt; /* the # of sit entries in set */ 309 }; 310 311 /* 312 * inline functions 313 */ 314 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type) 315 { 316 if (type == CURSEG_COLD_DATA_PINNED) 317 type = CURSEG_COLD_DATA; 318 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); 319 } 320 321 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, 322 unsigned int segno) 323 { 324 struct sit_info *sit_i = SIT_I(sbi); 325 return &sit_i->sentries[segno]; 326 } 327 328 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, 329 unsigned int segno) 330 { 331 struct sit_info *sit_i = SIT_I(sbi); 332 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; 333 } 334 335 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, 336 unsigned int segno, bool use_section) 337 { 338 /* 339 * In order to get # of valid blocks in a section instantly from many 340 * segments, f2fs manages two counting structures separately. 341 */ 342 if (use_section && __is_large_section(sbi)) 343 return get_sec_entry(sbi, segno)->valid_blocks; 344 else 345 return get_seg_entry(sbi, segno)->valid_blocks; 346 } 347 348 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi, 349 unsigned int segno) 350 { 351 return get_seg_entry(sbi, segno)->ckpt_valid_blocks; 352 } 353 354 static inline void seg_info_from_raw_sit(struct seg_entry *se, 355 struct f2fs_sit_entry *rs) 356 { 357 se->valid_blocks = GET_SIT_VBLOCKS(rs); 358 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); 359 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 360 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 361 #ifdef CONFIG_F2FS_CHECK_FS 362 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 363 #endif 364 se->type = GET_SIT_TYPE(rs); 365 se->mtime = le64_to_cpu(rs->mtime); 366 } 367 368 static inline void __seg_info_to_raw_sit(struct seg_entry *se, 369 struct f2fs_sit_entry *rs) 370 { 371 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | 372 se->valid_blocks; 373 rs->vblocks = cpu_to_le16(raw_vblocks); 374 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); 375 rs->mtime = cpu_to_le64(se->mtime); 376 } 377 378 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi, 379 struct page *page, unsigned int start) 380 { 381 struct f2fs_sit_block *raw_sit; 382 struct seg_entry *se; 383 struct f2fs_sit_entry *rs; 384 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK, 385 (unsigned long)MAIN_SEGS(sbi)); 386 int i; 387 388 raw_sit = (struct f2fs_sit_block *)page_address(page); 389 memset(raw_sit, 0, PAGE_SIZE); 390 for (i = 0; i < end - start; i++) { 391 rs = &raw_sit->entries[i]; 392 se = get_seg_entry(sbi, start + i); 393 __seg_info_to_raw_sit(se, rs); 394 } 395 } 396 397 static inline void seg_info_to_raw_sit(struct seg_entry *se, 398 struct f2fs_sit_entry *rs) 399 { 400 __seg_info_to_raw_sit(se, rs); 401 402 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 403 se->ckpt_valid_blocks = se->valid_blocks; 404 } 405 406 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, 407 unsigned int max, unsigned int segno) 408 { 409 unsigned int ret; 410 spin_lock(&free_i->segmap_lock); 411 ret = find_next_bit(free_i->free_segmap, max, segno); 412 spin_unlock(&free_i->segmap_lock); 413 return ret; 414 } 415 416 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) 417 { 418 struct free_segmap_info *free_i = FREE_I(sbi); 419 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 420 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 421 unsigned int next; 422 423 spin_lock(&free_i->segmap_lock); 424 clear_bit(segno, free_i->free_segmap); 425 free_i->free_segments++; 426 427 next = find_next_bit(free_i->free_segmap, 428 start_segno + sbi->segs_per_sec, start_segno); 429 if (next >= start_segno + sbi->segs_per_sec) { 430 clear_bit(secno, free_i->free_secmap); 431 free_i->free_sections++; 432 } 433 spin_unlock(&free_i->segmap_lock); 434 } 435 436 static inline void __set_inuse(struct f2fs_sb_info *sbi, 437 unsigned int segno) 438 { 439 struct free_segmap_info *free_i = FREE_I(sbi); 440 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 441 442 set_bit(segno, free_i->free_segmap); 443 free_i->free_segments--; 444 if (!test_and_set_bit(secno, free_i->free_secmap)) 445 free_i->free_sections--; 446 } 447 448 static inline void __set_test_and_free(struct f2fs_sb_info *sbi, 449 unsigned int segno) 450 { 451 struct free_segmap_info *free_i = FREE_I(sbi); 452 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 453 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 454 unsigned int next; 455 456 spin_lock(&free_i->segmap_lock); 457 if (test_and_clear_bit(segno, free_i->free_segmap)) { 458 free_i->free_segments++; 459 460 if (IS_CURSEC(sbi, secno)) 461 goto skip_free; 462 next = find_next_bit(free_i->free_segmap, 463 start_segno + sbi->segs_per_sec, start_segno); 464 if (next >= start_segno + sbi->segs_per_sec) { 465 if (test_and_clear_bit(secno, free_i->free_secmap)) 466 free_i->free_sections++; 467 } 468 } 469 skip_free: 470 spin_unlock(&free_i->segmap_lock); 471 } 472 473 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, 474 unsigned int segno) 475 { 476 struct free_segmap_info *free_i = FREE_I(sbi); 477 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 478 479 spin_lock(&free_i->segmap_lock); 480 if (!test_and_set_bit(segno, free_i->free_segmap)) { 481 free_i->free_segments--; 482 if (!test_and_set_bit(secno, free_i->free_secmap)) 483 free_i->free_sections--; 484 } 485 spin_unlock(&free_i->segmap_lock); 486 } 487 488 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, 489 void *dst_addr) 490 { 491 struct sit_info *sit_i = SIT_I(sbi); 492 493 #ifdef CONFIG_F2FS_CHECK_FS 494 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, 495 sit_i->bitmap_size)) 496 f2fs_bug_on(sbi, 1); 497 #endif 498 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); 499 } 500 501 static inline block_t written_block_count(struct f2fs_sb_info *sbi) 502 { 503 return SIT_I(sbi)->written_valid_blocks; 504 } 505 506 static inline unsigned int free_segments(struct f2fs_sb_info *sbi) 507 { 508 return FREE_I(sbi)->free_segments; 509 } 510 511 static inline int reserved_segments(struct f2fs_sb_info *sbi) 512 { 513 return SM_I(sbi)->reserved_segments; 514 } 515 516 static inline unsigned int free_sections(struct f2fs_sb_info *sbi) 517 { 518 return FREE_I(sbi)->free_sections; 519 } 520 521 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) 522 { 523 return DIRTY_I(sbi)->nr_dirty[PRE]; 524 } 525 526 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) 527 { 528 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + 529 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + 530 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + 531 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + 532 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + 533 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; 534 } 535 536 static inline int overprovision_segments(struct f2fs_sb_info *sbi) 537 { 538 return SM_I(sbi)->ovp_segments; 539 } 540 541 static inline int reserved_sections(struct f2fs_sb_info *sbi) 542 { 543 return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi)); 544 } 545 546 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi) 547 { 548 unsigned int node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) + 549 get_pages(sbi, F2FS_DIRTY_DENTS); 550 unsigned int dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS); 551 unsigned int segno, left_blocks; 552 int i; 553 554 /* check current node segment */ 555 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) { 556 segno = CURSEG_I(sbi, i)->segno; 557 left_blocks = sbi->blocks_per_seg - 558 get_seg_entry(sbi, segno)->ckpt_valid_blocks; 559 560 if (node_blocks > left_blocks) 561 return false; 562 } 563 564 /* check current data segment */ 565 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno; 566 left_blocks = sbi->blocks_per_seg - 567 get_seg_entry(sbi, segno)->ckpt_valid_blocks; 568 if (dent_blocks > left_blocks) 569 return false; 570 return true; 571 } 572 573 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, 574 int freed, int needed) 575 { 576 int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES); 577 int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS); 578 int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA); 579 580 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 581 return false; 582 583 if (free_sections(sbi) + freed == reserved_sections(sbi) + needed && 584 has_curseg_enough_space(sbi)) 585 return false; 586 return (free_sections(sbi) + freed) <= 587 (node_secs + 2 * dent_secs + imeta_secs + 588 reserved_sections(sbi) + needed); 589 } 590 591 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi) 592 { 593 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 594 return true; 595 if (likely(!has_not_enough_free_secs(sbi, 0, 0))) 596 return true; 597 return false; 598 } 599 600 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) 601 { 602 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; 603 } 604 605 static inline int utilization(struct f2fs_sb_info *sbi) 606 { 607 return div_u64((u64)valid_user_blocks(sbi) * 100, 608 sbi->user_block_count); 609 } 610 611 /* 612 * Sometimes f2fs may be better to drop out-of-place update policy. 613 * And, users can control the policy through sysfs entries. 614 * There are five policies with triggering conditions as follows. 615 * F2FS_IPU_FORCE - all the time, 616 * F2FS_IPU_SSR - if SSR mode is activated, 617 * F2FS_IPU_UTIL - if FS utilization is over threashold, 618 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over 619 * threashold, 620 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash 621 * storages. IPU will be triggered only if the # of dirty 622 * pages over min_fsync_blocks. 623 * F2FS_IPUT_DISABLE - disable IPU. (=default option) 624 */ 625 #define DEF_MIN_IPU_UTIL 70 626 #define DEF_MIN_FSYNC_BLOCKS 8 627 #define DEF_MIN_HOT_BLOCKS 16 628 629 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */ 630 631 enum { 632 F2FS_IPU_FORCE, 633 F2FS_IPU_SSR, 634 F2FS_IPU_UTIL, 635 F2FS_IPU_SSR_UTIL, 636 F2FS_IPU_FSYNC, 637 F2FS_IPU_ASYNC, 638 }; 639 640 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, 641 int type) 642 { 643 struct curseg_info *curseg = CURSEG_I(sbi, type); 644 return curseg->segno; 645 } 646 647 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, 648 int type) 649 { 650 struct curseg_info *curseg = CURSEG_I(sbi, type); 651 return curseg->alloc_type; 652 } 653 654 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type) 655 { 656 struct curseg_info *curseg = CURSEG_I(sbi, type); 657 return curseg->next_blkoff; 658 } 659 660 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno) 661 { 662 f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1); 663 } 664 665 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio) 666 { 667 struct f2fs_sb_info *sbi = fio->sbi; 668 669 if (__is_valid_data_blkaddr(fio->old_blkaddr)) 670 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ? 671 META_GENERIC : DATA_GENERIC); 672 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ? 673 META_GENERIC : DATA_GENERIC_ENHANCE); 674 } 675 676 /* 677 * Summary block is always treated as an invalid block 678 */ 679 static inline int check_block_count(struct f2fs_sb_info *sbi, 680 int segno, struct f2fs_sit_entry *raw_sit) 681 { 682 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; 683 int valid_blocks = 0; 684 int cur_pos = 0, next_pos; 685 686 /* check bitmap with valid block count */ 687 do { 688 if (is_valid) { 689 next_pos = find_next_zero_bit_le(&raw_sit->valid_map, 690 sbi->blocks_per_seg, 691 cur_pos); 692 valid_blocks += next_pos - cur_pos; 693 } else 694 next_pos = find_next_bit_le(&raw_sit->valid_map, 695 sbi->blocks_per_seg, 696 cur_pos); 697 cur_pos = next_pos; 698 is_valid = !is_valid; 699 } while (cur_pos < sbi->blocks_per_seg); 700 701 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) { 702 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d", 703 GET_SIT_VBLOCKS(raw_sit), valid_blocks); 704 set_sbi_flag(sbi, SBI_NEED_FSCK); 705 return -EFSCORRUPTED; 706 } 707 708 /* check segment usage, and check boundary of a given segment number */ 709 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg 710 || segno > TOTAL_SEGS(sbi) - 1)) { 711 f2fs_err(sbi, "Wrong valid blocks %d or segno %u", 712 GET_SIT_VBLOCKS(raw_sit), segno); 713 set_sbi_flag(sbi, SBI_NEED_FSCK); 714 return -EFSCORRUPTED; 715 } 716 return 0; 717 } 718 719 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, 720 unsigned int start) 721 { 722 struct sit_info *sit_i = SIT_I(sbi); 723 unsigned int offset = SIT_BLOCK_OFFSET(start); 724 block_t blk_addr = sit_i->sit_base_addr + offset; 725 726 check_seg_range(sbi, start); 727 728 #ifdef CONFIG_F2FS_CHECK_FS 729 if (f2fs_test_bit(offset, sit_i->sit_bitmap) != 730 f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) 731 f2fs_bug_on(sbi, 1); 732 #endif 733 734 /* calculate sit block address */ 735 if (f2fs_test_bit(offset, sit_i->sit_bitmap)) 736 blk_addr += sit_i->sit_blocks; 737 738 return blk_addr; 739 } 740 741 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, 742 pgoff_t block_addr) 743 { 744 struct sit_info *sit_i = SIT_I(sbi); 745 block_addr -= sit_i->sit_base_addr; 746 if (block_addr < sit_i->sit_blocks) 747 block_addr += sit_i->sit_blocks; 748 else 749 block_addr -= sit_i->sit_blocks; 750 751 return block_addr + sit_i->sit_base_addr; 752 } 753 754 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) 755 { 756 unsigned int block_off = SIT_BLOCK_OFFSET(start); 757 758 f2fs_change_bit(block_off, sit_i->sit_bitmap); 759 #ifdef CONFIG_F2FS_CHECK_FS 760 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); 761 #endif 762 } 763 764 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi, 765 bool base_time) 766 { 767 struct sit_info *sit_i = SIT_I(sbi); 768 time64_t diff, now = ktime_get_real_seconds(); 769 770 if (now >= sit_i->mounted_time) 771 return sit_i->elapsed_time + now - sit_i->mounted_time; 772 773 /* system time is set to the past */ 774 if (!base_time) { 775 diff = sit_i->mounted_time - now; 776 if (sit_i->elapsed_time >= diff) 777 return sit_i->elapsed_time - diff; 778 return 0; 779 } 780 return sit_i->elapsed_time; 781 } 782 783 static inline void set_summary(struct f2fs_summary *sum, nid_t nid, 784 unsigned int ofs_in_node, unsigned char version) 785 { 786 sum->nid = cpu_to_le32(nid); 787 sum->ofs_in_node = cpu_to_le16(ofs_in_node); 788 sum->version = version; 789 } 790 791 static inline block_t start_sum_block(struct f2fs_sb_info *sbi) 792 { 793 return __start_cp_addr(sbi) + 794 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); 795 } 796 797 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) 798 { 799 return __start_cp_addr(sbi) + 800 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) 801 - (base + 1) + type; 802 } 803 804 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) 805 { 806 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) 807 return true; 808 return false; 809 } 810 811 /* 812 * It is very important to gather dirty pages and write at once, so that we can 813 * submit a big bio without interfering other data writes. 814 * By default, 512 pages for directory data, 815 * 512 pages (2MB) * 8 for nodes, and 816 * 256 pages * 8 for meta are set. 817 */ 818 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) 819 { 820 if (sbi->sb->s_bdi->wb.dirty_exceeded) 821 return 0; 822 823 if (type == DATA) 824 return sbi->blocks_per_seg; 825 else if (type == NODE) 826 return 8 * sbi->blocks_per_seg; 827 else if (type == META) 828 return 8 * BIO_MAX_PAGES; 829 else 830 return 0; 831 } 832 833 /* 834 * When writing pages, it'd better align nr_to_write for segment size. 835 */ 836 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, 837 struct writeback_control *wbc) 838 { 839 long nr_to_write, desired; 840 841 if (wbc->sync_mode != WB_SYNC_NONE) 842 return 0; 843 844 nr_to_write = wbc->nr_to_write; 845 desired = BIO_MAX_PAGES; 846 if (type == NODE) 847 desired <<= 1; 848 849 wbc->nr_to_write = desired; 850 return desired - nr_to_write; 851 } 852 853 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force) 854 { 855 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info; 856 bool wakeup = false; 857 int i; 858 859 if (force) 860 goto wake_up; 861 862 mutex_lock(&dcc->cmd_lock); 863 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) { 864 if (i + 1 < dcc->discard_granularity) 865 break; 866 if (!list_empty(&dcc->pend_list[i])) { 867 wakeup = true; 868 break; 869 } 870 } 871 mutex_unlock(&dcc->cmd_lock); 872 if (!wakeup || !is_idle(sbi, DISCARD_TIME)) 873 return; 874 wake_up: 875 dcc->discard_wake = 1; 876 wake_up_interruptible_all(&dcc->discard_wait_queue); 877 } 878