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 }; 192 193 struct seg_entry { 194 unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */ 195 unsigned int valid_blocks:10; /* # of valid blocks */ 196 unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */ 197 unsigned int padding:6; /* padding */ 198 unsigned char *cur_valid_map; /* validity bitmap of blocks */ 199 #ifdef CONFIG_F2FS_CHECK_FS 200 unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */ 201 #endif 202 /* 203 * # of valid blocks and the validity bitmap stored in the last 204 * checkpoint pack. This information is used by the SSR mode. 205 */ 206 unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */ 207 unsigned char *discard_map; 208 unsigned long long mtime; /* modification time of the segment */ 209 }; 210 211 struct sec_entry { 212 unsigned int valid_blocks; /* # of valid blocks in a section */ 213 }; 214 215 #define MAX_SKIP_GC_COUNT 16 216 217 struct revoke_entry { 218 struct list_head list; 219 block_t old_addr; /* for revoking when fail to commit */ 220 pgoff_t index; 221 }; 222 223 struct sit_info { 224 block_t sit_base_addr; /* start block address of SIT area */ 225 block_t sit_blocks; /* # of blocks used by SIT area */ 226 block_t written_valid_blocks; /* # of valid blocks in main area */ 227 char *bitmap; /* all bitmaps pointer */ 228 char *sit_bitmap; /* SIT bitmap pointer */ 229 #ifdef CONFIG_F2FS_CHECK_FS 230 char *sit_bitmap_mir; /* SIT bitmap mirror */ 231 232 /* bitmap of segments to be ignored by GC in case of errors */ 233 unsigned long *invalid_segmap; 234 #endif 235 unsigned int bitmap_size; /* SIT bitmap size */ 236 237 unsigned long *tmp_map; /* bitmap for temporal use */ 238 unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */ 239 unsigned int dirty_sentries; /* # of dirty sentries */ 240 unsigned int sents_per_block; /* # of SIT entries per block */ 241 struct rw_semaphore sentry_lock; /* to protect SIT cache */ 242 struct seg_entry *sentries; /* SIT segment-level cache */ 243 struct sec_entry *sec_entries; /* SIT section-level cache */ 244 245 /* for cost-benefit algorithm in cleaning procedure */ 246 unsigned long long elapsed_time; /* elapsed time after mount */ 247 unsigned long long mounted_time; /* mount time */ 248 unsigned long long min_mtime; /* min. modification time */ 249 unsigned long long max_mtime; /* max. modification time */ 250 unsigned long long dirty_min_mtime; /* rerange candidates in GC_AT */ 251 unsigned long long dirty_max_mtime; /* rerange candidates in GC_AT */ 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 unsigned long *dirty_segmap[NR_DIRTY_TYPE]; 280 unsigned long *dirty_secmap; 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 unsigned long *pinned_secmap; /* pinned victims from foreground GC */ 285 unsigned int pinned_secmap_cnt; /* count of victims which has pinned data */ 286 bool enable_pin_section; /* enable pinning section */ 287 }; 288 289 /* for active log information */ 290 struct curseg_info { 291 struct mutex curseg_mutex; /* lock for consistency */ 292 struct f2fs_summary_block *sum_blk; /* cached summary block */ 293 struct rw_semaphore journal_rwsem; /* protect journal area */ 294 struct f2fs_journal *journal; /* cached journal info */ 295 unsigned char alloc_type; /* current allocation type */ 296 unsigned short seg_type; /* segment type like CURSEG_XXX_TYPE */ 297 unsigned int segno; /* current segment number */ 298 unsigned short next_blkoff; /* next block offset to write */ 299 unsigned int zone; /* current zone number */ 300 unsigned int next_segno; /* preallocated segment */ 301 int fragment_remained_chunk; /* remained block size in a chunk for block fragmentation mode */ 302 bool inited; /* indicate inmem log is inited */ 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 return (struct curseg_info *)(SM_I(sbi)->curseg_array + type); 317 } 318 319 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi, 320 unsigned int segno) 321 { 322 struct sit_info *sit_i = SIT_I(sbi); 323 return &sit_i->sentries[segno]; 324 } 325 326 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi, 327 unsigned int segno) 328 { 329 struct sit_info *sit_i = SIT_I(sbi); 330 return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)]; 331 } 332 333 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi, 334 unsigned int segno, bool use_section) 335 { 336 /* 337 * In order to get # of valid blocks in a section instantly from many 338 * segments, f2fs manages two counting structures separately. 339 */ 340 if (use_section && __is_large_section(sbi)) 341 return get_sec_entry(sbi, segno)->valid_blocks; 342 else 343 return get_seg_entry(sbi, segno)->valid_blocks; 344 } 345 346 static inline unsigned int get_ckpt_valid_blocks(struct f2fs_sb_info *sbi, 347 unsigned int segno, bool use_section) 348 { 349 if (use_section && __is_large_section(sbi)) { 350 unsigned int start_segno = START_SEGNO(segno); 351 unsigned int blocks = 0; 352 int i; 353 354 for (i = 0; i < SEGS_PER_SEC(sbi); i++, start_segno++) { 355 struct seg_entry *se = get_seg_entry(sbi, start_segno); 356 357 blocks += se->ckpt_valid_blocks; 358 } 359 return blocks; 360 } 361 return get_seg_entry(sbi, segno)->ckpt_valid_blocks; 362 } 363 364 static inline void seg_info_from_raw_sit(struct seg_entry *se, 365 struct f2fs_sit_entry *rs) 366 { 367 se->valid_blocks = GET_SIT_VBLOCKS(rs); 368 se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs); 369 memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 370 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 371 #ifdef CONFIG_F2FS_CHECK_FS 372 memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 373 #endif 374 se->type = GET_SIT_TYPE(rs); 375 se->mtime = le64_to_cpu(rs->mtime); 376 } 377 378 static inline void __seg_info_to_raw_sit(struct seg_entry *se, 379 struct f2fs_sit_entry *rs) 380 { 381 unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) | 382 se->valid_blocks; 383 rs->vblocks = cpu_to_le16(raw_vblocks); 384 memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE); 385 rs->mtime = cpu_to_le64(se->mtime); 386 } 387 388 static inline void seg_info_to_sit_page(struct f2fs_sb_info *sbi, 389 struct page *page, unsigned int start) 390 { 391 struct f2fs_sit_block *raw_sit; 392 struct seg_entry *se; 393 struct f2fs_sit_entry *rs; 394 unsigned int end = min(start + SIT_ENTRY_PER_BLOCK, 395 (unsigned long)MAIN_SEGS(sbi)); 396 int i; 397 398 raw_sit = (struct f2fs_sit_block *)page_address(page); 399 memset(raw_sit, 0, PAGE_SIZE); 400 for (i = 0; i < end - start; i++) { 401 rs = &raw_sit->entries[i]; 402 se = get_seg_entry(sbi, start + i); 403 __seg_info_to_raw_sit(se, rs); 404 } 405 } 406 407 static inline void seg_info_to_raw_sit(struct seg_entry *se, 408 struct f2fs_sit_entry *rs) 409 { 410 __seg_info_to_raw_sit(se, rs); 411 412 memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE); 413 se->ckpt_valid_blocks = se->valid_blocks; 414 } 415 416 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i, 417 unsigned int max, unsigned int segno) 418 { 419 unsigned int ret; 420 spin_lock(&free_i->segmap_lock); 421 ret = find_next_bit(free_i->free_segmap, max, segno); 422 spin_unlock(&free_i->segmap_lock); 423 return ret; 424 } 425 426 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno) 427 { 428 struct free_segmap_info *free_i = FREE_I(sbi); 429 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 430 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 431 unsigned int next; 432 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno); 433 434 spin_lock(&free_i->segmap_lock); 435 clear_bit(segno, free_i->free_segmap); 436 free_i->free_segments++; 437 438 next = find_next_bit(free_i->free_segmap, 439 start_segno + SEGS_PER_SEC(sbi), start_segno); 440 if (next >= start_segno + usable_segs) { 441 clear_bit(secno, free_i->free_secmap); 442 free_i->free_sections++; 443 } 444 spin_unlock(&free_i->segmap_lock); 445 } 446 447 static inline void __set_inuse(struct f2fs_sb_info *sbi, 448 unsigned int segno) 449 { 450 struct free_segmap_info *free_i = FREE_I(sbi); 451 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 452 453 set_bit(segno, free_i->free_segmap); 454 free_i->free_segments--; 455 if (!test_and_set_bit(secno, free_i->free_secmap)) 456 free_i->free_sections--; 457 } 458 459 static inline void __set_test_and_free(struct f2fs_sb_info *sbi, 460 unsigned int segno, bool inmem) 461 { 462 struct free_segmap_info *free_i = FREE_I(sbi); 463 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 464 unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno); 465 unsigned int next; 466 unsigned int usable_segs = f2fs_usable_segs_in_sec(sbi, segno); 467 468 spin_lock(&free_i->segmap_lock); 469 if (test_and_clear_bit(segno, free_i->free_segmap)) { 470 free_i->free_segments++; 471 472 if (!inmem && IS_CURSEC(sbi, secno)) 473 goto skip_free; 474 next = find_next_bit(free_i->free_segmap, 475 start_segno + SEGS_PER_SEC(sbi), start_segno); 476 if (next >= start_segno + usable_segs) { 477 if (test_and_clear_bit(secno, free_i->free_secmap)) 478 free_i->free_sections++; 479 } 480 } 481 skip_free: 482 spin_unlock(&free_i->segmap_lock); 483 } 484 485 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi, 486 unsigned int segno) 487 { 488 struct free_segmap_info *free_i = FREE_I(sbi); 489 unsigned int secno = GET_SEC_FROM_SEG(sbi, segno); 490 491 spin_lock(&free_i->segmap_lock); 492 if (!test_and_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 spin_unlock(&free_i->segmap_lock); 498 } 499 500 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi, 501 void *dst_addr) 502 { 503 struct sit_info *sit_i = SIT_I(sbi); 504 505 #ifdef CONFIG_F2FS_CHECK_FS 506 if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir, 507 sit_i->bitmap_size)) 508 f2fs_bug_on(sbi, 1); 509 #endif 510 memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size); 511 } 512 513 static inline block_t written_block_count(struct f2fs_sb_info *sbi) 514 { 515 return SIT_I(sbi)->written_valid_blocks; 516 } 517 518 static inline unsigned int free_segments(struct f2fs_sb_info *sbi) 519 { 520 return FREE_I(sbi)->free_segments; 521 } 522 523 static inline unsigned int reserved_segments(struct f2fs_sb_info *sbi) 524 { 525 return SM_I(sbi)->reserved_segments + 526 SM_I(sbi)->additional_reserved_segments; 527 } 528 529 static inline unsigned int free_sections(struct f2fs_sb_info *sbi) 530 { 531 return FREE_I(sbi)->free_sections; 532 } 533 534 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi) 535 { 536 return DIRTY_I(sbi)->nr_dirty[PRE]; 537 } 538 539 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi) 540 { 541 return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] + 542 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] + 543 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] + 544 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] + 545 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] + 546 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE]; 547 } 548 549 static inline int overprovision_segments(struct f2fs_sb_info *sbi) 550 { 551 return SM_I(sbi)->ovp_segments; 552 } 553 554 static inline int reserved_sections(struct f2fs_sb_info *sbi) 555 { 556 return GET_SEC_FROM_SEG(sbi, reserved_segments(sbi)); 557 } 558 559 static inline bool has_curseg_enough_space(struct f2fs_sb_info *sbi, 560 unsigned int node_blocks, unsigned int dent_blocks) 561 { 562 563 unsigned segno, left_blocks; 564 int i; 565 566 /* check current node sections in the worst case. */ 567 for (i = CURSEG_HOT_NODE; i <= CURSEG_COLD_NODE; i++) { 568 segno = CURSEG_I(sbi, i)->segno; 569 left_blocks = CAP_BLKS_PER_SEC(sbi) - 570 get_ckpt_valid_blocks(sbi, segno, true); 571 if (node_blocks > left_blocks) 572 return false; 573 } 574 575 /* check current data section for dentry blocks. */ 576 segno = CURSEG_I(sbi, CURSEG_HOT_DATA)->segno; 577 left_blocks = CAP_BLKS_PER_SEC(sbi) - 578 get_ckpt_valid_blocks(sbi, segno, true); 579 if (dent_blocks > left_blocks) 580 return false; 581 return true; 582 } 583 584 /* 585 * calculate needed sections for dirty node/dentry 586 * and call has_curseg_enough_space 587 */ 588 static inline void __get_secs_required(struct f2fs_sb_info *sbi, 589 unsigned int *lower_p, unsigned int *upper_p, bool *curseg_p) 590 { 591 unsigned int total_node_blocks = get_pages(sbi, F2FS_DIRTY_NODES) + 592 get_pages(sbi, F2FS_DIRTY_DENTS) + 593 get_pages(sbi, F2FS_DIRTY_IMETA); 594 unsigned int total_dent_blocks = get_pages(sbi, F2FS_DIRTY_DENTS); 595 unsigned int node_secs = total_node_blocks / CAP_BLKS_PER_SEC(sbi); 596 unsigned int dent_secs = total_dent_blocks / CAP_BLKS_PER_SEC(sbi); 597 unsigned int node_blocks = total_node_blocks % CAP_BLKS_PER_SEC(sbi); 598 unsigned int dent_blocks = total_dent_blocks % CAP_BLKS_PER_SEC(sbi); 599 600 if (lower_p) 601 *lower_p = node_secs + dent_secs; 602 if (upper_p) 603 *upper_p = node_secs + dent_secs + 604 (node_blocks ? 1 : 0) + (dent_blocks ? 1 : 0); 605 if (curseg_p) 606 *curseg_p = has_curseg_enough_space(sbi, 607 node_blocks, dent_blocks); 608 } 609 610 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, 611 int freed, int needed) 612 { 613 unsigned int free_secs, lower_secs, upper_secs; 614 bool curseg_space; 615 616 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 617 return false; 618 619 __get_secs_required(sbi, &lower_secs, &upper_secs, &curseg_space); 620 621 free_secs = free_sections(sbi) + freed; 622 lower_secs += needed + reserved_sections(sbi); 623 upper_secs += needed + reserved_sections(sbi); 624 625 if (free_secs > upper_secs) 626 return false; 627 if (free_secs <= lower_secs) 628 return true; 629 return !curseg_space; 630 } 631 632 static inline bool has_enough_free_secs(struct f2fs_sb_info *sbi, 633 int freed, int needed) 634 { 635 return !has_not_enough_free_secs(sbi, freed, needed); 636 } 637 638 static inline bool f2fs_is_checkpoint_ready(struct f2fs_sb_info *sbi) 639 { 640 if (likely(!is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 641 return true; 642 if (likely(has_enough_free_secs(sbi, 0, 0))) 643 return true; 644 return false; 645 } 646 647 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi) 648 { 649 return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments; 650 } 651 652 static inline int utilization(struct f2fs_sb_info *sbi) 653 { 654 return div_u64((u64)valid_user_blocks(sbi) * 100, 655 sbi->user_block_count); 656 } 657 658 /* 659 * Sometimes f2fs may be better to drop out-of-place update policy. 660 * And, users can control the policy through sysfs entries. 661 * There are five policies with triggering conditions as follows. 662 * F2FS_IPU_FORCE - all the time, 663 * F2FS_IPU_SSR - if SSR mode is activated, 664 * F2FS_IPU_UTIL - if FS utilization is over threashold, 665 * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over 666 * threashold, 667 * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash 668 * storages. IPU will be triggered only if the # of dirty 669 * pages over min_fsync_blocks. (=default option) 670 * F2FS_IPU_ASYNC - do IPU given by asynchronous write requests. 671 * F2FS_IPU_NOCACHE - disable IPU bio cache. 672 * F2FS_IPU_HONOR_OPU_WRITE - use OPU write prior to IPU write if inode has 673 * FI_OPU_WRITE flag. 674 * F2FS_IPU_DISABLE - disable IPU. (=default option in LFS mode) 675 */ 676 #define DEF_MIN_IPU_UTIL 70 677 #define DEF_MIN_FSYNC_BLOCKS 8 678 #define DEF_MIN_HOT_BLOCKS 16 679 680 #define SMALL_VOLUME_SEGMENTS (16 * 512) /* 16GB */ 681 682 #define F2FS_IPU_DISABLE 0 683 684 /* Modification on enum should be synchronized with ipu_mode_names array */ 685 enum { 686 F2FS_IPU_FORCE, 687 F2FS_IPU_SSR, 688 F2FS_IPU_UTIL, 689 F2FS_IPU_SSR_UTIL, 690 F2FS_IPU_FSYNC, 691 F2FS_IPU_ASYNC, 692 F2FS_IPU_NOCACHE, 693 F2FS_IPU_HONOR_OPU_WRITE, 694 F2FS_IPU_MAX, 695 }; 696 697 static inline bool IS_F2FS_IPU_DISABLE(struct f2fs_sb_info *sbi) 698 { 699 return SM_I(sbi)->ipu_policy == F2FS_IPU_DISABLE; 700 } 701 702 #define F2FS_IPU_POLICY(name) \ 703 static inline bool IS_##name(struct f2fs_sb_info *sbi) \ 704 { \ 705 return SM_I(sbi)->ipu_policy & BIT(name); \ 706 } 707 708 F2FS_IPU_POLICY(F2FS_IPU_FORCE); 709 F2FS_IPU_POLICY(F2FS_IPU_SSR); 710 F2FS_IPU_POLICY(F2FS_IPU_UTIL); 711 F2FS_IPU_POLICY(F2FS_IPU_SSR_UTIL); 712 F2FS_IPU_POLICY(F2FS_IPU_FSYNC); 713 F2FS_IPU_POLICY(F2FS_IPU_ASYNC); 714 F2FS_IPU_POLICY(F2FS_IPU_NOCACHE); 715 F2FS_IPU_POLICY(F2FS_IPU_HONOR_OPU_WRITE); 716 717 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi, 718 int type) 719 { 720 struct curseg_info *curseg = CURSEG_I(sbi, type); 721 return curseg->segno; 722 } 723 724 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi, 725 int type) 726 { 727 struct curseg_info *curseg = CURSEG_I(sbi, type); 728 return curseg->alloc_type; 729 } 730 731 static inline bool valid_main_segno(struct f2fs_sb_info *sbi, 732 unsigned int segno) 733 { 734 return segno <= (MAIN_SEGS(sbi) - 1); 735 } 736 737 static inline void verify_fio_blkaddr(struct f2fs_io_info *fio) 738 { 739 struct f2fs_sb_info *sbi = fio->sbi; 740 741 if (__is_valid_data_blkaddr(fio->old_blkaddr)) 742 verify_blkaddr(sbi, fio->old_blkaddr, __is_meta_io(fio) ? 743 META_GENERIC : DATA_GENERIC); 744 verify_blkaddr(sbi, fio->new_blkaddr, __is_meta_io(fio) ? 745 META_GENERIC : DATA_GENERIC_ENHANCE); 746 } 747 748 /* 749 * Summary block is always treated as an invalid block 750 */ 751 static inline int check_block_count(struct f2fs_sb_info *sbi, 752 int segno, struct f2fs_sit_entry *raw_sit) 753 { 754 bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false; 755 int valid_blocks = 0; 756 int cur_pos = 0, next_pos; 757 unsigned int usable_blks_per_seg = f2fs_usable_blks_in_seg(sbi, segno); 758 759 /* check bitmap with valid block count */ 760 do { 761 if (is_valid) { 762 next_pos = find_next_zero_bit_le(&raw_sit->valid_map, 763 usable_blks_per_seg, 764 cur_pos); 765 valid_blocks += next_pos - cur_pos; 766 } else 767 next_pos = find_next_bit_le(&raw_sit->valid_map, 768 usable_blks_per_seg, 769 cur_pos); 770 cur_pos = next_pos; 771 is_valid = !is_valid; 772 } while (cur_pos < usable_blks_per_seg); 773 774 if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) { 775 f2fs_err(sbi, "Mismatch valid blocks %d vs. %d", 776 GET_SIT_VBLOCKS(raw_sit), valid_blocks); 777 set_sbi_flag(sbi, SBI_NEED_FSCK); 778 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 779 return -EFSCORRUPTED; 780 } 781 782 if (usable_blks_per_seg < BLKS_PER_SEG(sbi)) 783 f2fs_bug_on(sbi, find_next_bit_le(&raw_sit->valid_map, 784 BLKS_PER_SEG(sbi), 785 usable_blks_per_seg) != BLKS_PER_SEG(sbi)); 786 787 /* check segment usage, and check boundary of a given segment number */ 788 if (unlikely(GET_SIT_VBLOCKS(raw_sit) > usable_blks_per_seg 789 || !valid_main_segno(sbi, segno))) { 790 f2fs_err(sbi, "Wrong valid blocks %d or segno %u", 791 GET_SIT_VBLOCKS(raw_sit), segno); 792 set_sbi_flag(sbi, SBI_NEED_FSCK); 793 f2fs_handle_error(sbi, ERROR_INCONSISTENT_SIT); 794 return -EFSCORRUPTED; 795 } 796 return 0; 797 } 798 799 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi, 800 unsigned int start) 801 { 802 struct sit_info *sit_i = SIT_I(sbi); 803 unsigned int offset = SIT_BLOCK_OFFSET(start); 804 block_t blk_addr = sit_i->sit_base_addr + offset; 805 806 f2fs_bug_on(sbi, !valid_main_segno(sbi, start)); 807 808 #ifdef CONFIG_F2FS_CHECK_FS 809 if (f2fs_test_bit(offset, sit_i->sit_bitmap) != 810 f2fs_test_bit(offset, sit_i->sit_bitmap_mir)) 811 f2fs_bug_on(sbi, 1); 812 #endif 813 814 /* calculate sit block address */ 815 if (f2fs_test_bit(offset, sit_i->sit_bitmap)) 816 blk_addr += sit_i->sit_blocks; 817 818 return blk_addr; 819 } 820 821 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi, 822 pgoff_t block_addr) 823 { 824 struct sit_info *sit_i = SIT_I(sbi); 825 block_addr -= sit_i->sit_base_addr; 826 if (block_addr < sit_i->sit_blocks) 827 block_addr += sit_i->sit_blocks; 828 else 829 block_addr -= sit_i->sit_blocks; 830 831 return block_addr + sit_i->sit_base_addr; 832 } 833 834 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start) 835 { 836 unsigned int block_off = SIT_BLOCK_OFFSET(start); 837 838 f2fs_change_bit(block_off, sit_i->sit_bitmap); 839 #ifdef CONFIG_F2FS_CHECK_FS 840 f2fs_change_bit(block_off, sit_i->sit_bitmap_mir); 841 #endif 842 } 843 844 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi, 845 bool base_time) 846 { 847 struct sit_info *sit_i = SIT_I(sbi); 848 time64_t diff, now = ktime_get_boottime_seconds(); 849 850 if (now >= sit_i->mounted_time) 851 return sit_i->elapsed_time + now - sit_i->mounted_time; 852 853 /* system time is set to the past */ 854 if (!base_time) { 855 diff = sit_i->mounted_time - now; 856 if (sit_i->elapsed_time >= diff) 857 return sit_i->elapsed_time - diff; 858 return 0; 859 } 860 return sit_i->elapsed_time; 861 } 862 863 static inline void set_summary(struct f2fs_summary *sum, nid_t nid, 864 unsigned int ofs_in_node, unsigned char version) 865 { 866 sum->nid = cpu_to_le32(nid); 867 sum->ofs_in_node = cpu_to_le16(ofs_in_node); 868 sum->version = version; 869 } 870 871 static inline block_t start_sum_block(struct f2fs_sb_info *sbi) 872 { 873 return __start_cp_addr(sbi) + 874 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum); 875 } 876 877 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type) 878 { 879 return __start_cp_addr(sbi) + 880 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count) 881 - (base + 1) + type; 882 } 883 884 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno) 885 { 886 if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno)) 887 return true; 888 return false; 889 } 890 891 /* 892 * It is very important to gather dirty pages and write at once, so that we can 893 * submit a big bio without interfering other data writes. 894 * By default, 512 pages for directory data, 895 * 512 pages (2MB) * 8 for nodes, and 896 * 256 pages * 8 for meta are set. 897 */ 898 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type) 899 { 900 if (sbi->sb->s_bdi->wb.dirty_exceeded) 901 return 0; 902 903 if (type == DATA) 904 return BLKS_PER_SEG(sbi); 905 else if (type == NODE) 906 return SEGS_TO_BLKS(sbi, 8); 907 else if (type == META) 908 return 8 * BIO_MAX_VECS; 909 else 910 return 0; 911 } 912 913 /* 914 * When writing pages, it'd better align nr_to_write for segment size. 915 */ 916 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type, 917 struct writeback_control *wbc) 918 { 919 long nr_to_write, desired; 920 921 if (wbc->sync_mode != WB_SYNC_NONE) 922 return 0; 923 924 nr_to_write = wbc->nr_to_write; 925 desired = BIO_MAX_VECS; 926 if (type == NODE) 927 desired <<= 1; 928 929 wbc->nr_to_write = desired; 930 return desired - nr_to_write; 931 } 932 933 static inline void wake_up_discard_thread(struct f2fs_sb_info *sbi, bool force) 934 { 935 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info; 936 bool wakeup = false; 937 int i; 938 939 if (force) 940 goto wake_up; 941 942 mutex_lock(&dcc->cmd_lock); 943 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) { 944 if (i + 1 < dcc->discard_granularity) 945 break; 946 if (!list_empty(&dcc->pend_list[i])) { 947 wakeup = true; 948 break; 949 } 950 } 951 mutex_unlock(&dcc->cmd_lock); 952 if (!wakeup || !is_idle(sbi, DISCARD_TIME)) 953 return; 954 wake_up: 955 dcc->discard_wake = true; 956 wake_up_interruptible_all(&dcc->discard_wait_queue); 957 } 958 959 static inline unsigned int first_zoned_segno(struct f2fs_sb_info *sbi) 960 { 961 int devi; 962 963 for (devi = 0; devi < sbi->s_ndevs; devi++) 964 if (bdev_is_zoned(FDEV(devi).bdev)) 965 return GET_SEGNO(sbi, FDEV(devi).start_blk); 966 return 0; 967 } 968