fs/f2fs/segment.h

   1 /*
   2  * fs/f2fs/segment.h
   3  *
   4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
   5  *             http://www.samsung.com/
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include <linux/blkdev.h>
  12 #include <linux/backing-dev.h>
  13
  14 /* constant macro */
  15 #define NULL_SEGNO                      ((unsigned int)(~0))
  16 #define NULL_SECNO                      ((unsigned int)(~0))
  17
  18 #define DEF_RECLAIM_PREFREE_SEGMENTS    5       /* 5% over total segments */
  19 #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS        4096    /* 8GB in maximum */
  20
  21 #define F2FS_MIN_SEGMENTS       9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
  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)
  29
  30 #define IS_HOT(t)       ((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
  31 #define IS_WARM(t)      ((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
  32 #define IS_COLD(t)      ((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
  33
  34 #define IS_CURSEG(sbi, seg)                                             \
  35         (((seg) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) ||    \
  36          ((seg) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) ||   \
  37          ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) ||   \
  38          ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) ||    \
  39          ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) ||   \
  40          ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
  41
  42 #define IS_CURSEC(sbi, secno)                                           \
  43         (((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno /            \
  44           (sbi)->segs_per_sec) ||       \
  45          ((secno) == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno /           \
  46           (sbi)->segs_per_sec) ||       \
  47          ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno /           \
  48           (sbi)->segs_per_sec) ||       \
  49          ((secno) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno /            \
  50           (sbi)->segs_per_sec) ||       \
  51          ((secno) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno /           \
  52           (sbi)->segs_per_sec) ||       \
  53          ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno /           \
  54           (sbi)->segs_per_sec)) \
  55
  56 #define MAIN_BLKADDR(sbi)       (SM_I(sbi)->main_blkaddr)
  57 #define SEG0_BLKADDR(sbi)       (SM_I(sbi)->seg0_blkaddr)
  58
  59 #define MAIN_SEGS(sbi)  (SM_I(sbi)->main_segments)
  60 #define MAIN_SECS(sbi)  ((sbi)->total_sections)
  61
  62 #define TOTAL_SEGS(sbi) (SM_I(sbi)->segment_count)
  63 #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << (sbi)->log_blocks_per_seg)
  64
  65 #define MAX_BLKADDR(sbi)        (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
  66 #define SEGMENT_SIZE(sbi)       (1ULL << ((sbi)->log_blocksize +        \
  67                                         (sbi)->log_blocks_per_seg))
  68
  69 #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) +                    \
  70          (GET_R2L_SEGNO(FREE_I(sbi), segno) << (sbi)->log_blocks_per_seg))
  71
  72 #define NEXT_FREE_BLKADDR(sbi, curseg)                                  \
  73         (START_BLOCK(sbi, (curseg)->segno) + (curseg)->next_blkoff)
  74
  75 #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr)     ((blk_addr) - SEG0_BLKADDR(sbi))
  76 #define GET_SEGNO_FROM_SEG0(sbi, blk_addr)                              \
  77         (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> (sbi)->log_blocks_per_seg)
  78 #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr)                             \
  79         (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & ((sbi)->blocks_per_seg - 1))
  80
  81 #define GET_SEGNO(sbi, blk_addr)                                        \
  82         ((((blk_addr) == NULL_ADDR) || ((blk_addr) == NEW_ADDR)) ?      \
  83         NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi),                 \
  84                 GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
  85 #define BLKS_PER_SEC(sbi)                                       \
  86         ((sbi)->segs_per_sec * (sbi)->blocks_per_seg)
  87 #define GET_SEC_FROM_SEG(sbi, segno)                            \
  88         ((segno) / (sbi)->segs_per_sec)
  89 #define GET_SEG_FROM_SEC(sbi, secno)                            \
  90         ((secno) * (sbi)->segs_per_sec)
  91 #define GET_ZONE_FROM_SEC(sbi, secno)                           \
  92         ((secno) / (sbi)->secs_per_zone)
  93 #define GET_ZONE_FROM_SEG(sbi, segno)                           \
  94         GET_ZONE_FROM_SEC(sbi, GET_SEC_FROM_SEG(sbi, segno))
  95
  96 #define GET_SUM_BLOCK(sbi, segno)                               \
  97         ((sbi)->sm_info->ssa_blkaddr + (segno))
  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         ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / 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  * indicate a block allocation direction: RIGHT and LEFT.
 120  * RIGHT means allocating new sections towards the end of volume.
 121  * LEFT means the opposite direction.
 122  */
 123 enum {
 124         ALLOC_RIGHT = 0,
 125         ALLOC_LEFT
 126 };
 127
 128 /*
 129  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
 130  * LFS writes data sequentially with cleaning operations.
 131  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
 132  */
 133 enum {
 134         LFS = 0,
 135         SSR
 136 };
 137
 138 /*
 139  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
 140  * GC_CB is based on cost-benefit algorithm.
 141  * GC_GREEDY is based on greedy algorithm.
 142  */
 143 enum {
 144         GC_CB = 0,
 145         GC_GREEDY,
 146         ALLOC_NEXT,
 147         FLUSH_DEVICE,
 148         MAX_GC_POLICY,
 149 };
 150
 151 /*
 152  * BG_GC means the background cleaning job.
 153  * FG_GC means the on-demand cleaning job.
 154  * FORCE_FG_GC means on-demand cleaning job in background.
 155  */
 156 enum {
 157         BG_GC = 0,
 158         FG_GC,
 159         FORCE_FG_GC,
 160 };
 161
 162 /* for a function parameter to select a victim segment */
 163 struct victim_sel_policy {
 164         int alloc_mode;                 /* LFS or SSR */
 165         int gc_mode;                    /* GC_CB or GC_GREEDY */
 166         unsigned long *dirty_segmap;    /* dirty segment bitmap */
 167         unsigned int max_search;        /* maximum # of segments to search */
 168         unsigned int offset;            /* last scanned bitmap offset */
 169         unsigned int ofs_unit;          /* bitmap search unit */
 170         unsigned int min_cost;          /* minimum cost */
 171         unsigned int min_segno;         /* segment # having min. cost */
 172 };
 173
 174 struct 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 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
 192 struct sec_entry {
 193         unsigned int valid_blocks;      /* # of valid blocks in a section */
 194 };
 195
 196 struct segment_allocation {
 197         void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
 198 };
 199
 200 /*
 201  * this value is set in page as a private data which indicate that
 202  * the page is atomically written, and it is in inmem_pages list.
 203  */
 204 #define ATOMIC_WRITTEN_PAGE             ((unsigned long)-1)
 205 #define DUMMY_WRITTEN_PAGE              ((unsigned long)-2)
 206
 207 #define IS_ATOMIC_WRITTEN_PAGE(page)                    \
 208                 (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
 209 #define IS_DUMMY_WRITTEN_PAGE(page)                     \
 210                 (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
 211
 212 struct inmem_pages {
 213         struct list_head list;
 214         struct page *page;
 215         block_t old_addr;               /* for revoking when fail to commit */
 216 };
 217
 218 struct sit_info {
 219         const struct segment_allocation *s_ops;
 220
 221         block_t sit_base_addr;          /* start block address of SIT area */
 222         block_t sit_blocks;             /* # of blocks used by SIT area */
 223         block_t written_valid_blocks;   /* # of valid blocks in main area */
 224         char *sit_bitmap;               /* SIT bitmap pointer */
 225 #ifdef CONFIG_F2FS_CHECK_FS
 226         char *sit_bitmap_mir;           /* SIT bitmap mirror */
 227 #endif
 228         unsigned int bitmap_size;       /* SIT bitmap size */
 229
 230         unsigned long *tmp_map;                 /* bitmap for temporal use */
 231         unsigned long *dirty_sentries_bitmap;   /* bitmap for dirty sentries */
 232         unsigned int dirty_sentries;            /* # of dirty sentries */
 233         unsigned int sents_per_block;           /* # of SIT entries per block */
 234         struct mutex sentry_lock;               /* to protect SIT cache */
 235         struct seg_entry *sentries;             /* SIT segment-level cache */
 236         struct sec_entry *sec_entries;          /* SIT section-level cache */
 237
 238         /* for cost-benefit algorithm in cleaning procedure */
 239         unsigned long long elapsed_time;        /* elapsed time after mount */
 240         unsigned long long mounted_time;        /* mount time */
 241         unsigned long long min_mtime;           /* min. modification time */
 242         unsigned long long max_mtime;           /* max. modification time */
 243
 244         unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
 245 };
 246
 247 struct free_segmap_info {
 248         unsigned int start_segno;       /* start segment number logically */
 249         unsigned int free_segments;     /* # of free segments */
 250         unsigned int free_sections;     /* # of free sections */
 251         spinlock_t segmap_lock;         /* free segmap lock */
 252         unsigned long *free_segmap;     /* free segment bitmap */
 253         unsigned long *free_secmap;     /* free section bitmap */
 254 };
 255
 256 /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
 257 enum dirty_type {
 258         DIRTY_HOT_DATA,         /* dirty segments assigned as hot data logs */
 259         DIRTY_WARM_DATA,        /* dirty segments assigned as warm data logs */
 260         DIRTY_COLD_DATA,        /* dirty segments assigned as cold data logs */
 261         DIRTY_HOT_NODE,         /* dirty segments assigned as hot node logs */
 262         DIRTY_WARM_NODE,        /* dirty segments assigned as warm node logs */
 263         DIRTY_COLD_NODE,        /* dirty segments assigned as cold node logs */
 264         DIRTY,                  /* to count # of dirty segments */
 265         PRE,                    /* to count # of entirely obsolete segments */
 266         NR_DIRTY_TYPE
 267 };
 268
 269 struct dirty_seglist_info {
 270         const struct victim_selection *v_ops;   /* victim selction operation */
 271         unsigned long *dirty_segmap[NR_DIRTY_TYPE];
 272         struct mutex seglist_lock;              /* lock for segment bitmaps */
 273         int nr_dirty[NR_DIRTY_TYPE];            /* # of dirty segments */
 274         unsigned long *victim_secmap;           /* background GC victims */
 275 };
 276
 277 /* victim selection function for cleaning and SSR */
 278 struct victim_selection {
 279         int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
 280                                                         int, int, char);
 281 };
 282
 283 /* for active log information */
 284 struct curseg_info {
 285         struct mutex curseg_mutex;              /* lock for consistency */
 286         struct f2fs_summary_block *sum_blk;     /* cached summary block */
 287         struct rw_semaphore journal_rwsem;      /* protect journal area */
 288         struct f2fs_journal *journal;           /* cached journal info */
 289         unsigned char alloc_type;               /* current allocation type */
 290         unsigned int segno;                     /* current segment number */
 291         unsigned short next_blkoff;             /* next block offset to write */
 292         unsigned int zone;                      /* current zone number */
 293         unsigned int next_segno;                /* preallocated segment */
 294 };
 295
 296 struct sit_entry_set {
 297         struct list_head set_list;      /* link with all sit sets */
 298         unsigned int start_segno;       /* start segno of sits in set */
 299         unsigned int entry_cnt;         /* the # of sit entries in set */
 300 };
 301
 302 /*
 303  * inline functions
 304  */
 305 static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
 306 {
 307         return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
 308 }
 309
 310 static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
 311                                                 unsigned int segno)
 312 {
 313         struct sit_info *sit_i = SIT_I(sbi);
 314         return &sit_i->sentries[segno];
 315 }
 316
 317 static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
 318                                                 unsigned int segno)
 319 {
 320         struct sit_info *sit_i = SIT_I(sbi);
 321         return &sit_i->sec_entries[GET_SEC_FROM_SEG(sbi, segno)];
 322 }
 323
 324 static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
 325                                 unsigned int segno, bool use_section)
 326 {
 327         /*
 328          * In order to get # of valid blocks in a section instantly from many
 329          * segments, f2fs manages two counting structures separately.
 330          */
 331         if (use_section && sbi->segs_per_sec > 1)
 332                 return get_sec_entry(sbi, segno)->valid_blocks;
 333         else
 334                 return get_seg_entry(sbi, segno)->valid_blocks;
 335 }
 336
 337 static inline void seg_info_from_raw_sit(struct seg_entry *se,
 338                                         struct f2fs_sit_entry *rs)
 339 {
 340         se->valid_blocks = GET_SIT_VBLOCKS(rs);
 341         se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
 342         memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
 343         memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
 344 #ifdef CONFIG_F2FS_CHECK_FS
 345         memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
 346 #endif
 347         se->type = GET_SIT_TYPE(rs);
 348         se->mtime = le64_to_cpu(rs->mtime);
 349 }
 350
 351 static inline void seg_info_to_raw_sit(struct seg_entry *se,
 352                                         struct f2fs_sit_entry *rs)
 353 {
 354         unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
 355                                         se->valid_blocks;
 356         rs->vblocks = cpu_to_le16(raw_vblocks);
 357         memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
 358         memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
 359         se->ckpt_valid_blocks = se->valid_blocks;
 360         rs->mtime = cpu_to_le64(se->mtime);
 361 }
 362
 363 static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
 364                 unsigned int max, unsigned int segno)
 365 {
 366         unsigned int ret;
 367         spin_lock(&free_i->segmap_lock);
 368         ret = find_next_bit(free_i->free_segmap, max, segno);
 369         spin_unlock(&free_i->segmap_lock);
 370         return ret;
 371 }
 372
 373 static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
 374 {
 375         struct free_segmap_info *free_i = FREE_I(sbi);
 376         unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
 377         unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
 378         unsigned int next;
 379
 380         spin_lock(&free_i->segmap_lock);
 381         clear_bit(segno, free_i->free_segmap);
 382         free_i->free_segments++;
 383
 384         next = find_next_bit(free_i->free_segmap,
 385                         start_segno + sbi->segs_per_sec, start_segno);
 386         if (next >= start_segno + sbi->segs_per_sec) {
 387                 clear_bit(secno, free_i->free_secmap);
 388                 free_i->free_sections++;
 389         }
 390         spin_unlock(&free_i->segmap_lock);
 391 }
 392
 393 static inline void __set_inuse(struct f2fs_sb_info *sbi,
 394                 unsigned int segno)
 395 {
 396         struct free_segmap_info *free_i = FREE_I(sbi);
 397         unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
 398
 399         set_bit(segno, free_i->free_segmap);
 400         free_i->free_segments--;
 401         if (!test_and_set_bit(secno, free_i->free_secmap))
 402                 free_i->free_sections--;
 403 }
 404
 405 static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
 406                 unsigned int segno)
 407 {
 408         struct free_segmap_info *free_i = FREE_I(sbi);
 409         unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
 410         unsigned int start_segno = GET_SEG_FROM_SEC(sbi, secno);
 411         unsigned int next;
 412
 413         spin_lock(&free_i->segmap_lock);
 414         if (test_and_clear_bit(segno, free_i->free_segmap)) {
 415                 free_i->free_segments++;
 416
 417                 next = find_next_bit(free_i->free_segmap,
 418                                 start_segno + sbi->segs_per_sec, start_segno);
 419                 if (next >= start_segno + sbi->segs_per_sec) {
 420                         if (test_and_clear_bit(secno, free_i->free_secmap))
 421                                 free_i->free_sections++;
 422                 }
 423         }
 424         spin_unlock(&free_i->segmap_lock);
 425 }
 426
 427 static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
 428                 unsigned int segno)
 429 {
 430         struct free_segmap_info *free_i = FREE_I(sbi);
 431         unsigned int secno = GET_SEC_FROM_SEG(sbi, segno);
 432
 433         spin_lock(&free_i->segmap_lock);
 434         if (!test_and_set_bit(segno, free_i->free_segmap)) {
 435                 free_i->free_segments--;
 436                 if (!test_and_set_bit(secno, free_i->free_secmap))
 437                         free_i->free_sections--;
 438         }
 439         spin_unlock(&free_i->segmap_lock);
 440 }
 441
 442 static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
 443                 void *dst_addr)
 444 {
 445         struct sit_info *sit_i = SIT_I(sbi);
 446
 447 #ifdef CONFIG_F2FS_CHECK_FS
 448         if (memcmp(sit_i->sit_bitmap, sit_i->sit_bitmap_mir,
 449                                                 sit_i->bitmap_size))
 450                 f2fs_bug_on(sbi, 1);
 451 #endif
 452         memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
 453 }
 454
 455 static inline block_t written_block_count(struct f2fs_sb_info *sbi)
 456 {
 457         return SIT_I(sbi)->written_valid_blocks;
 458 }
 459
 460 static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
 461 {
 462         return FREE_I(sbi)->free_segments;
 463 }
 464
 465 static inline int reserved_segments(struct f2fs_sb_info *sbi)
 466 {
 467         return SM_I(sbi)->reserved_segments;
 468 }
 469
 470 static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
 471 {
 472         return FREE_I(sbi)->free_sections;
 473 }
 474
 475 static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
 476 {
 477         return DIRTY_I(sbi)->nr_dirty[PRE];
 478 }
 479
 480 static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
 481 {
 482         return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
 483                 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
 484                 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
 485                 DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
 486                 DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
 487                 DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
 488 }
 489
 490 static inline int overprovision_segments(struct f2fs_sb_info *sbi)
 491 {
 492         return SM_I(sbi)->ovp_segments;
 493 }
 494
 495 static inline int overprovision_sections(struct f2fs_sb_info *sbi)
 496 {
 497         return GET_SEC_FROM_SEG(sbi, (unsigned int)overprovision_segments(sbi));
 498 }
 499
 500 static inline int reserved_sections(struct f2fs_sb_info *sbi)
 501 {
 502         return GET_SEC_FROM_SEG(sbi, (unsigned int)reserved_segments(sbi));
 503 }
 504
 505 static inline bool need_SSR(struct f2fs_sb_info *sbi)
 506 {
 507         int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
 508         int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
 509         int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
 510
 511         if (test_opt(sbi, LFS))
 512                 return false;
 513
 514         return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
 515                                                 2 * reserved_sections(sbi));
 516 }
 517
 518 static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
 519                                         int freed, int needed)
 520 {
 521         int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
 522         int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
 523         int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
 524
 525         if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
 526                 return false;
 527
 528         return (free_sections(sbi) + freed) <=
 529                 (node_secs + 2 * dent_secs + imeta_secs +
 530                 reserved_sections(sbi) + needed);
 531 }
 532
 533 static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
 534 {
 535         return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
 536 }
 537
 538 static inline int utilization(struct f2fs_sb_info *sbi)
 539 {
 540         return div_u64((u64)valid_user_blocks(sbi) * 100,
 541                                         sbi->user_block_count);
 542 }
 543
 544 /*
 545  * Sometimes f2fs may be better to drop out-of-place update policy.
 546  * And, users can control the policy through sysfs entries.
 547  * There are five policies with triggering conditions as follows.
 548  * F2FS_IPU_FORCE - all the time,
 549  * F2FS_IPU_SSR - if SSR mode is activated,
 550  * F2FS_IPU_UTIL - if FS utilization is over threashold,
 551  * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
 552  *                     threashold,
 553  * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
 554  *                     storages. IPU will be triggered only if the # of dirty
 555  *                     pages over min_fsync_blocks.
 556  * F2FS_IPUT_DISABLE - disable IPU. (=default option)
 557  */
 558 #define DEF_MIN_IPU_UTIL        70
 559 #define DEF_MIN_FSYNC_BLOCKS    8
 560 #define DEF_MIN_HOT_BLOCKS      16
 561
 562 enum {
 563         F2FS_IPU_FORCE,
 564         F2FS_IPU_SSR,
 565         F2FS_IPU_UTIL,
 566         F2FS_IPU_SSR_UTIL,
 567         F2FS_IPU_FSYNC,
 568         F2FS_IPU_ASYNC,
 569 };
 570
 571 static inline bool need_inplace_update_policy(struct inode *inode,
 572                                 struct f2fs_io_info *fio)
 573 {
 574         struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 575         unsigned int policy = SM_I(sbi)->ipu_policy;
 576
 577         if (test_opt(sbi, LFS))
 578                 return false;
 579
 580         if (policy & (0x1 << F2FS_IPU_FORCE))
 581                 return true;
 582         if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi))
 583                 return true;
 584         if (policy & (0x1 << F2FS_IPU_UTIL) &&
 585                         utilization(sbi) > SM_I(sbi)->min_ipu_util)
 586                 return true;
 587         if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) &&
 588                         utilization(sbi) > SM_I(sbi)->min_ipu_util)
 589                 return true;
 590
 591         /*
 592          * IPU for rewrite async pages
 593          */
 594         if (policy & (0x1 << F2FS_IPU_ASYNC) &&
 595                         fio && fio->op == REQ_OP_WRITE &&
 596                         !(fio->op_flags & REQ_SYNC) &&
 597                         !f2fs_encrypted_inode(inode))
 598                 return true;
 599
 600         /* this is only set during fdatasync */
 601         if (policy & (0x1 << F2FS_IPU_FSYNC) &&
 602                         is_inode_flag_set(inode, FI_NEED_IPU))
 603                 return true;
 604
 605         return false;
 606 }
 607
 608 static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
 609                 int type)
 610 {
 611         struct curseg_info *curseg = CURSEG_I(sbi, type);
 612         return curseg->segno;
 613 }
 614
 615 static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
 616                 int type)
 617 {
 618         struct curseg_info *curseg = CURSEG_I(sbi, type);
 619         return curseg->alloc_type;
 620 }
 621
 622 static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
 623 {
 624         struct curseg_info *curseg = CURSEG_I(sbi, type);
 625         return curseg->next_blkoff;
 626 }
 627
 628 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
 629 {
 630         f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
 631 }
 632
 633 static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
 634 {
 635         BUG_ON(blk_addr < SEG0_BLKADDR(sbi)
 636                         || blk_addr >= MAX_BLKADDR(sbi));
 637 }
 638
 639 /*
 640  * Summary block is always treated as an invalid block
 641  */
 642 static inline void check_block_count(struct f2fs_sb_info *sbi,
 643                 int segno, struct f2fs_sit_entry *raw_sit)
 644 {
 645 #ifdef CONFIG_F2FS_CHECK_FS
 646         bool is_valid  = test_bit_le(0, raw_sit->valid_map) ? true : false;
 647         int valid_blocks = 0;
 648         int cur_pos = 0, next_pos;
 649
 650         /* check bitmap with valid block count */
 651         do {
 652                 if (is_valid) {
 653                         next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
 654                                         sbi->blocks_per_seg,
 655                                         cur_pos);
 656                         valid_blocks += next_pos - cur_pos;
 657                 } else
 658                         next_pos = find_next_bit_le(&raw_sit->valid_map,
 659                                         sbi->blocks_per_seg,
 660                                         cur_pos);
 661                 cur_pos = next_pos;
 662                 is_valid = !is_valid;
 663         } while (cur_pos < sbi->blocks_per_seg);
 664         BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
 665 #endif
 666         /* check segment usage, and check boundary of a given segment number */
 667         f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
 668                                         || segno > TOTAL_SEGS(sbi) - 1);
 669 }
 670
 671 static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
 672                                                 unsigned int start)
 673 {
 674         struct sit_info *sit_i = SIT_I(sbi);
 675         unsigned int offset = SIT_BLOCK_OFFSET(start);
 676         block_t blk_addr = sit_i->sit_base_addr + offset;
 677
 678         check_seg_range(sbi, start);
 679
 680 #ifdef CONFIG_F2FS_CHECK_FS
 681         if (f2fs_test_bit(offset, sit_i->sit_bitmap) !=
 682                         f2fs_test_bit(offset, sit_i->sit_bitmap_mir))
 683                 f2fs_bug_on(sbi, 1);
 684 #endif
 685
 686         /* calculate sit block address */
 687         if (f2fs_test_bit(offset, sit_i->sit_bitmap))
 688                 blk_addr += sit_i->sit_blocks;
 689
 690         return blk_addr;
 691 }
 692
 693 static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
 694                                                 pgoff_t block_addr)
 695 {
 696         struct sit_info *sit_i = SIT_I(sbi);
 697         block_addr -= sit_i->sit_base_addr;
 698         if (block_addr < sit_i->sit_blocks)
 699                 block_addr += sit_i->sit_blocks;
 700         else
 701                 block_addr -= sit_i->sit_blocks;
 702
 703         return block_addr + sit_i->sit_base_addr;
 704 }
 705
 706 static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
 707 {
 708         unsigned int block_off = SIT_BLOCK_OFFSET(start);
 709
 710         f2fs_change_bit(block_off, sit_i->sit_bitmap);
 711 #ifdef CONFIG_F2FS_CHECK_FS
 712         f2fs_change_bit(block_off, sit_i->sit_bitmap_mir);
 713 #endif
 714 }
 715
 716 static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
 717 {
 718         struct sit_info *sit_i = SIT_I(sbi);
 719         time64_t now = ktime_get_real_seconds();
 720
 721         return sit_i->elapsed_time + now - sit_i->mounted_time;
 722 }
 723
 724 static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
 725                         unsigned int ofs_in_node, unsigned char version)
 726 {
 727         sum->nid = cpu_to_le32(nid);
 728         sum->ofs_in_node = cpu_to_le16(ofs_in_node);
 729         sum->version = version;
 730 }
 731
 732 static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
 733 {
 734         return __start_cp_addr(sbi) +
 735                 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
 736 }
 737
 738 static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
 739 {
 740         return __start_cp_addr(sbi) +
 741                 le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
 742                                 - (base + 1) + type;
 743 }
 744
 745 static inline bool no_fggc_candidate(struct f2fs_sb_info *sbi,
 746                                                 unsigned int secno)
 747 {
 748         if (get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno), true) >=
 749                                                 sbi->fggc_threshold)
 750                 return true;
 751         return false;
 752 }
 753
 754 static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
 755 {
 756         if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
 757                 return true;
 758         return false;
 759 }
 760
 761 /*
 762  * It is very important to gather dirty pages and write at once, so that we can
 763  * submit a big bio without interfering other data writes.
 764  * By default, 512 pages for directory data,
 765  * 512 pages (2MB) * 8 for nodes, and
 766  * 256 pages * 8 for meta are set.
 767  */
 768 static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
 769 {
 770         if (sbi->sb->s_bdi->wb.dirty_exceeded)
 771                 return 0;
 772
 773         if (type == DATA)
 774                 return sbi->blocks_per_seg;
 775         else if (type == NODE)
 776                 return 8 * sbi->blocks_per_seg;
 777         else if (type == META)
 778                 return 8 * BIO_MAX_PAGES;
 779         else
 780                 return 0;
 781 }
 782
 783 /*
 784  * When writing pages, it'd better align nr_to_write for segment size.
 785  */
 786 static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
 787                                         struct writeback_control *wbc)
 788 {
 789         long nr_to_write, desired;
 790
 791         if (wbc->sync_mode != WB_SYNC_NONE)
 792                 return 0;
 793
 794         nr_to_write = wbc->nr_to_write;
 795         desired = BIO_MAX_PAGES;
 796         if (type == NODE)
 797                 desired <<= 1;
 798
 799         wbc->nr_to_write = desired;
 800         return desired - nr_to_write;
 801 }