segment.h 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748
  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. /* constant macro */
  14. #define NULL_SEGNO ((unsigned int)(~0))
  15. #define NULL_SECNO ((unsigned int)(~0))
  16. #define DEF_RECLAIM_PREFREE_SEGMENTS 5 /* 5% over total segments */
  17. #define DEF_MAX_RECLAIM_PREFREE_SEGMENTS 4096 /* 8GB in maximum */
  18. #define F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
  19. /* L: Logical segment # in volume, R: Relative segment # in main area */
  20. #define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno)
  21. #define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno)
  22. #define IS_DATASEG(t) (t <= CURSEG_COLD_DATA)
  23. #define IS_NODESEG(t) (t >= CURSEG_HOT_NODE)
  24. #define IS_CURSEG(sbi, seg) \
  25. ((seg == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
  26. (seg == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
  27. (seg == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
  28. (seg == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
  29. (seg == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
  30. (seg == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
  31. #define IS_CURSEC(sbi, secno) \
  32. ((secno == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
  33. sbi->segs_per_sec) || \
  34. (secno == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
  35. sbi->segs_per_sec) || \
  36. (secno == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
  37. sbi->segs_per_sec) || \
  38. (secno == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
  39. sbi->segs_per_sec) || \
  40. (secno == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
  41. sbi->segs_per_sec) || \
  42. (secno == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
  43. sbi->segs_per_sec)) \
  44. #define MAIN_BLKADDR(sbi) (SM_I(sbi)->main_blkaddr)
  45. #define SEG0_BLKADDR(sbi) (SM_I(sbi)->seg0_blkaddr)
  46. #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
  47. #define MAIN_SECS(sbi) (sbi->total_sections)
  48. #define TOTAL_SEGS(sbi) (SM_I(sbi)->segment_count)
  49. #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << sbi->log_blocks_per_seg)
  50. #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
  51. #define SEGMENT_SIZE(sbi) (1ULL << (sbi->log_blocksize + \
  52. sbi->log_blocks_per_seg))
  53. #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
  54. (GET_R2L_SEGNO(FREE_I(sbi), segno) << sbi->log_blocks_per_seg))
  55. #define NEXT_FREE_BLKADDR(sbi, curseg) \
  56. (START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff)
  57. #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
  58. #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
  59. (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> sbi->log_blocks_per_seg)
  60. #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
  61. (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (sbi->blocks_per_seg - 1))
  62. #define GET_SEGNO(sbi, blk_addr) \
  63. (((blk_addr == NULL_ADDR) || (blk_addr == NEW_ADDR)) ? \
  64. NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
  65. GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
  66. #define GET_SECNO(sbi, segno) \
  67. ((segno) / sbi->segs_per_sec)
  68. #define GET_ZONENO_FROM_SEGNO(sbi, segno) \
  69. ((segno / sbi->segs_per_sec) / sbi->secs_per_zone)
  70. #define GET_SUM_BLOCK(sbi, segno) \
  71. ((sbi->sm_info->ssa_blkaddr) + segno)
  72. #define GET_SUM_TYPE(footer) ((footer)->entry_type)
  73. #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = type)
  74. #define SIT_ENTRY_OFFSET(sit_i, segno) \
  75. (segno % sit_i->sents_per_block)
  76. #define SIT_BLOCK_OFFSET(segno) \
  77. (segno / SIT_ENTRY_PER_BLOCK)
  78. #define START_SEGNO(segno) \
  79. (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
  80. #define SIT_BLK_CNT(sbi) \
  81. ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
  82. #define f2fs_bitmap_size(nr) \
  83. (BITS_TO_LONGS(nr) * sizeof(unsigned long))
  84. #define SECTOR_FROM_BLOCK(blk_addr) \
  85. (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
  86. #define SECTOR_TO_BLOCK(sectors) \
  87. (sectors >> F2FS_LOG_SECTORS_PER_BLOCK)
  88. /*
  89. * indicate a block allocation direction: RIGHT and LEFT.
  90. * RIGHT means allocating new sections towards the end of volume.
  91. * LEFT means the opposite direction.
  92. */
  93. enum {
  94. ALLOC_RIGHT = 0,
  95. ALLOC_LEFT
  96. };
  97. /*
  98. * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
  99. * LFS writes data sequentially with cleaning operations.
  100. * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
  101. */
  102. enum {
  103. LFS = 0,
  104. SSR
  105. };
  106. /*
  107. * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
  108. * GC_CB is based on cost-benefit algorithm.
  109. * GC_GREEDY is based on greedy algorithm.
  110. */
  111. enum {
  112. GC_CB = 0,
  113. GC_GREEDY
  114. };
  115. /*
  116. * BG_GC means the background cleaning job.
  117. * FG_GC means the on-demand cleaning job.
  118. * FORCE_FG_GC means on-demand cleaning job in background.
  119. */
  120. enum {
  121. BG_GC = 0,
  122. FG_GC,
  123. FORCE_FG_GC,
  124. };
  125. /* for a function parameter to select a victim segment */
  126. struct victim_sel_policy {
  127. int alloc_mode; /* LFS or SSR */
  128. int gc_mode; /* GC_CB or GC_GREEDY */
  129. unsigned long *dirty_segmap; /* dirty segment bitmap */
  130. unsigned int max_search; /* maximum # of segments to search */
  131. unsigned int offset; /* last scanned bitmap offset */
  132. unsigned int ofs_unit; /* bitmap search unit */
  133. unsigned int min_cost; /* minimum cost */
  134. unsigned int min_segno; /* segment # having min. cost */
  135. };
  136. struct seg_entry {
  137. unsigned int type:6; /* segment type like CURSEG_XXX_TYPE */
  138. unsigned int valid_blocks:10; /* # of valid blocks */
  139. unsigned int ckpt_valid_blocks:10; /* # of valid blocks last cp */
  140. unsigned int padding:6; /* padding */
  141. unsigned char *cur_valid_map; /* validity bitmap of blocks */
  142. #ifdef CONFIG_F2FS_CHECK_FS
  143. unsigned char *cur_valid_map_mir; /* mirror of current valid bitmap */
  144. #endif
  145. /*
  146. * # of valid blocks and the validity bitmap stored in the the last
  147. * checkpoint pack. This information is used by the SSR mode.
  148. */
  149. unsigned char *ckpt_valid_map; /* validity bitmap of blocks last cp */
  150. unsigned char *discard_map;
  151. unsigned long long mtime; /* modification time of the segment */
  152. };
  153. struct sec_entry {
  154. unsigned int valid_blocks; /* # of valid blocks in a section */
  155. };
  156. struct segment_allocation {
  157. void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
  158. };
  159. /*
  160. * this value is set in page as a private data which indicate that
  161. * the page is atomically written, and it is in inmem_pages list.
  162. */
  163. #define ATOMIC_WRITTEN_PAGE ((unsigned long)-1)
  164. #define DUMMY_WRITTEN_PAGE ((unsigned long)-2)
  165. #define IS_ATOMIC_WRITTEN_PAGE(page) \
  166. (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
  167. #define IS_DUMMY_WRITTEN_PAGE(page) \
  168. (page_private(page) == (unsigned long)DUMMY_WRITTEN_PAGE)
  169. struct inmem_pages {
  170. struct list_head list;
  171. struct page *page;
  172. block_t old_addr; /* for revoking when fail to commit */
  173. };
  174. struct sit_info {
  175. const struct segment_allocation *s_ops;
  176. block_t sit_base_addr; /* start block address of SIT area */
  177. block_t sit_blocks; /* # of blocks used by SIT area */
  178. block_t written_valid_blocks; /* # of valid blocks in main area */
  179. char *sit_bitmap; /* SIT bitmap pointer */
  180. unsigned int bitmap_size; /* SIT bitmap size */
  181. unsigned long *tmp_map; /* bitmap for temporal use */
  182. unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
  183. unsigned int dirty_sentries; /* # of dirty sentries */
  184. unsigned int sents_per_block; /* # of SIT entries per block */
  185. struct mutex sentry_lock; /* to protect SIT cache */
  186. struct seg_entry *sentries; /* SIT segment-level cache */
  187. struct sec_entry *sec_entries; /* SIT section-level cache */
  188. /* for cost-benefit algorithm in cleaning procedure */
  189. unsigned long long elapsed_time; /* elapsed time after mount */
  190. unsigned long long mounted_time; /* mount time */
  191. unsigned long long min_mtime; /* min. modification time */
  192. unsigned long long max_mtime; /* max. modification time */
  193. };
  194. struct free_segmap_info {
  195. unsigned int start_segno; /* start segment number logically */
  196. unsigned int free_segments; /* # of free segments */
  197. unsigned int free_sections; /* # of free sections */
  198. spinlock_t segmap_lock; /* free segmap lock */
  199. unsigned long *free_segmap; /* free segment bitmap */
  200. unsigned long *free_secmap; /* free section bitmap */
  201. };
  202. /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
  203. enum dirty_type {
  204. DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
  205. DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
  206. DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
  207. DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
  208. DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
  209. DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
  210. DIRTY, /* to count # of dirty segments */
  211. PRE, /* to count # of entirely obsolete segments */
  212. NR_DIRTY_TYPE
  213. };
  214. struct dirty_seglist_info {
  215. const struct victim_selection *v_ops; /* victim selction operation */
  216. unsigned long *dirty_segmap[NR_DIRTY_TYPE];
  217. struct mutex seglist_lock; /* lock for segment bitmaps */
  218. int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
  219. unsigned long *victim_secmap; /* background GC victims */
  220. };
  221. /* victim selection function for cleaning and SSR */
  222. struct victim_selection {
  223. int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
  224. int, int, char);
  225. };
  226. /* for active log information */
  227. struct curseg_info {
  228. struct mutex curseg_mutex; /* lock for consistency */
  229. struct f2fs_summary_block *sum_blk; /* cached summary block */
  230. struct rw_semaphore journal_rwsem; /* protect journal area */
  231. struct f2fs_journal *journal; /* cached journal info */
  232. unsigned char alloc_type; /* current allocation type */
  233. unsigned int segno; /* current segment number */
  234. unsigned short next_blkoff; /* next block offset to write */
  235. unsigned int zone; /* current zone number */
  236. unsigned int next_segno; /* preallocated segment */
  237. };
  238. struct sit_entry_set {
  239. struct list_head set_list; /* link with all sit sets */
  240. unsigned int start_segno; /* start segno of sits in set */
  241. unsigned int entry_cnt; /* the # of sit entries in set */
  242. };
  243. /*
  244. * inline functions
  245. */
  246. static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
  247. {
  248. return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
  249. }
  250. static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
  251. unsigned int segno)
  252. {
  253. struct sit_info *sit_i = SIT_I(sbi);
  254. return &sit_i->sentries[segno];
  255. }
  256. static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
  257. unsigned int segno)
  258. {
  259. struct sit_info *sit_i = SIT_I(sbi);
  260. return &sit_i->sec_entries[GET_SECNO(sbi, segno)];
  261. }
  262. static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
  263. unsigned int segno, int section)
  264. {
  265. /*
  266. * In order to get # of valid blocks in a section instantly from many
  267. * segments, f2fs manages two counting structures separately.
  268. */
  269. if (section > 1)
  270. return get_sec_entry(sbi, segno)->valid_blocks;
  271. else
  272. return get_seg_entry(sbi, segno)->valid_blocks;
  273. }
  274. static inline void seg_info_from_raw_sit(struct seg_entry *se,
  275. struct f2fs_sit_entry *rs)
  276. {
  277. se->valid_blocks = GET_SIT_VBLOCKS(rs);
  278. se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
  279. memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  280. memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  281. #ifdef CONFIG_F2FS_CHECK_FS
  282. memcpy(se->cur_valid_map_mir, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  283. #endif
  284. se->type = GET_SIT_TYPE(rs);
  285. se->mtime = le64_to_cpu(rs->mtime);
  286. }
  287. static inline void seg_info_to_raw_sit(struct seg_entry *se,
  288. struct f2fs_sit_entry *rs)
  289. {
  290. unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
  291. se->valid_blocks;
  292. rs->vblocks = cpu_to_le16(raw_vblocks);
  293. memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
  294. memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  295. se->ckpt_valid_blocks = se->valid_blocks;
  296. rs->mtime = cpu_to_le64(se->mtime);
  297. }
  298. static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
  299. unsigned int max, unsigned int segno)
  300. {
  301. unsigned int ret;
  302. spin_lock(&free_i->segmap_lock);
  303. ret = find_next_bit(free_i->free_segmap, max, segno);
  304. spin_unlock(&free_i->segmap_lock);
  305. return ret;
  306. }
  307. static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
  308. {
  309. struct free_segmap_info *free_i = FREE_I(sbi);
  310. unsigned int secno = segno / sbi->segs_per_sec;
  311. unsigned int start_segno = secno * sbi->segs_per_sec;
  312. unsigned int next;
  313. spin_lock(&free_i->segmap_lock);
  314. clear_bit(segno, free_i->free_segmap);
  315. free_i->free_segments++;
  316. next = find_next_bit(free_i->free_segmap,
  317. start_segno + sbi->segs_per_sec, start_segno);
  318. if (next >= start_segno + sbi->segs_per_sec) {
  319. clear_bit(secno, free_i->free_secmap);
  320. free_i->free_sections++;
  321. }
  322. spin_unlock(&free_i->segmap_lock);
  323. }
  324. static inline void __set_inuse(struct f2fs_sb_info *sbi,
  325. unsigned int segno)
  326. {
  327. struct free_segmap_info *free_i = FREE_I(sbi);
  328. unsigned int secno = segno / sbi->segs_per_sec;
  329. set_bit(segno, free_i->free_segmap);
  330. free_i->free_segments--;
  331. if (!test_and_set_bit(secno, free_i->free_secmap))
  332. free_i->free_sections--;
  333. }
  334. static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
  335. unsigned int segno)
  336. {
  337. struct free_segmap_info *free_i = FREE_I(sbi);
  338. unsigned int secno = segno / sbi->segs_per_sec;
  339. unsigned int start_segno = secno * sbi->segs_per_sec;
  340. unsigned int next;
  341. spin_lock(&free_i->segmap_lock);
  342. if (test_and_clear_bit(segno, free_i->free_segmap)) {
  343. free_i->free_segments++;
  344. next = find_next_bit(free_i->free_segmap,
  345. start_segno + sbi->segs_per_sec, start_segno);
  346. if (next >= start_segno + sbi->segs_per_sec) {
  347. if (test_and_clear_bit(secno, free_i->free_secmap))
  348. free_i->free_sections++;
  349. }
  350. }
  351. spin_unlock(&free_i->segmap_lock);
  352. }
  353. static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
  354. unsigned int segno)
  355. {
  356. struct free_segmap_info *free_i = FREE_I(sbi);
  357. unsigned int secno = segno / sbi->segs_per_sec;
  358. spin_lock(&free_i->segmap_lock);
  359. if (!test_and_set_bit(segno, free_i->free_segmap)) {
  360. free_i->free_segments--;
  361. if (!test_and_set_bit(secno, free_i->free_secmap))
  362. free_i->free_sections--;
  363. }
  364. spin_unlock(&free_i->segmap_lock);
  365. }
  366. static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
  367. void *dst_addr)
  368. {
  369. struct sit_info *sit_i = SIT_I(sbi);
  370. memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
  371. }
  372. static inline block_t written_block_count(struct f2fs_sb_info *sbi)
  373. {
  374. return SIT_I(sbi)->written_valid_blocks;
  375. }
  376. static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
  377. {
  378. return FREE_I(sbi)->free_segments;
  379. }
  380. static inline int reserved_segments(struct f2fs_sb_info *sbi)
  381. {
  382. return SM_I(sbi)->reserved_segments;
  383. }
  384. static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
  385. {
  386. return FREE_I(sbi)->free_sections;
  387. }
  388. static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
  389. {
  390. return DIRTY_I(sbi)->nr_dirty[PRE];
  391. }
  392. static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
  393. {
  394. return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
  395. DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
  396. DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
  397. DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
  398. DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
  399. DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
  400. }
  401. static inline int overprovision_segments(struct f2fs_sb_info *sbi)
  402. {
  403. return SM_I(sbi)->ovp_segments;
  404. }
  405. static inline int overprovision_sections(struct f2fs_sb_info *sbi)
  406. {
  407. return ((unsigned int) overprovision_segments(sbi)) / sbi->segs_per_sec;
  408. }
  409. static inline int reserved_sections(struct f2fs_sb_info *sbi)
  410. {
  411. return ((unsigned int) reserved_segments(sbi)) / sbi->segs_per_sec;
  412. }
  413. static inline bool need_SSR(struct f2fs_sb_info *sbi)
  414. {
  415. int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
  416. int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
  417. int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
  418. if (test_opt(sbi, LFS))
  419. return false;
  420. return free_sections(sbi) <= (node_secs + 2 * dent_secs + imeta_secs +
  421. reserved_sections(sbi) + 1);
  422. }
  423. static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi,
  424. int freed, int needed)
  425. {
  426. int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
  427. int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
  428. int imeta_secs = get_blocktype_secs(sbi, F2FS_DIRTY_IMETA);
  429. if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
  430. return false;
  431. return (free_sections(sbi) + freed) <=
  432. (node_secs + 2 * dent_secs + imeta_secs +
  433. reserved_sections(sbi) + needed);
  434. }
  435. static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
  436. {
  437. return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
  438. }
  439. static inline int utilization(struct f2fs_sb_info *sbi)
  440. {
  441. return div_u64((u64)valid_user_blocks(sbi) * 100,
  442. sbi->user_block_count);
  443. }
  444. /*
  445. * Sometimes f2fs may be better to drop out-of-place update policy.
  446. * And, users can control the policy through sysfs entries.
  447. * There are five policies with triggering conditions as follows.
  448. * F2FS_IPU_FORCE - all the time,
  449. * F2FS_IPU_SSR - if SSR mode is activated,
  450. * F2FS_IPU_UTIL - if FS utilization is over threashold,
  451. * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
  452. * threashold,
  453. * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
  454. * storages. IPU will be triggered only if the # of dirty
  455. * pages over min_fsync_blocks.
  456. * F2FS_IPUT_DISABLE - disable IPU. (=default option)
  457. */
  458. #define DEF_MIN_IPU_UTIL 70
  459. #define DEF_MIN_FSYNC_BLOCKS 8
  460. enum {
  461. F2FS_IPU_FORCE,
  462. F2FS_IPU_SSR,
  463. F2FS_IPU_UTIL,
  464. F2FS_IPU_SSR_UTIL,
  465. F2FS_IPU_FSYNC,
  466. };
  467. static inline bool need_inplace_update(struct inode *inode)
  468. {
  469. struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  470. unsigned int policy = SM_I(sbi)->ipu_policy;
  471. /* IPU can be done only for the user data */
  472. if (S_ISDIR(inode->i_mode) || f2fs_is_atomic_file(inode))
  473. return false;
  474. if (test_opt(sbi, LFS))
  475. return false;
  476. if (policy & (0x1 << F2FS_IPU_FORCE))
  477. return true;
  478. if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi))
  479. return true;
  480. if (policy & (0x1 << F2FS_IPU_UTIL) &&
  481. utilization(sbi) > SM_I(sbi)->min_ipu_util)
  482. return true;
  483. if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) &&
  484. utilization(sbi) > SM_I(sbi)->min_ipu_util)
  485. return true;
  486. /* this is only set during fdatasync */
  487. if (policy & (0x1 << F2FS_IPU_FSYNC) &&
  488. is_inode_flag_set(inode, FI_NEED_IPU))
  489. return true;
  490. return false;
  491. }
  492. static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
  493. int type)
  494. {
  495. struct curseg_info *curseg = CURSEG_I(sbi, type);
  496. return curseg->segno;
  497. }
  498. static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
  499. int type)
  500. {
  501. struct curseg_info *curseg = CURSEG_I(sbi, type);
  502. return curseg->alloc_type;
  503. }
  504. static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
  505. {
  506. struct curseg_info *curseg = CURSEG_I(sbi, type);
  507. return curseg->next_blkoff;
  508. }
  509. static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
  510. {
  511. f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
  512. }
  513. static inline void verify_block_addr(struct f2fs_sb_info *sbi, block_t blk_addr)
  514. {
  515. BUG_ON(blk_addr < SEG0_BLKADDR(sbi)
  516. || blk_addr >= MAX_BLKADDR(sbi));
  517. }
  518. /*
  519. * Summary block is always treated as an invalid block
  520. */
  521. static inline void check_block_count(struct f2fs_sb_info *sbi,
  522. int segno, struct f2fs_sit_entry *raw_sit)
  523. {
  524. #ifdef CONFIG_F2FS_CHECK_FS
  525. bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
  526. int valid_blocks = 0;
  527. int cur_pos = 0, next_pos;
  528. /* check bitmap with valid block count */
  529. do {
  530. if (is_valid) {
  531. next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
  532. sbi->blocks_per_seg,
  533. cur_pos);
  534. valid_blocks += next_pos - cur_pos;
  535. } else
  536. next_pos = find_next_bit_le(&raw_sit->valid_map,
  537. sbi->blocks_per_seg,
  538. cur_pos);
  539. cur_pos = next_pos;
  540. is_valid = !is_valid;
  541. } while (cur_pos < sbi->blocks_per_seg);
  542. BUG_ON(GET_SIT_VBLOCKS(raw_sit) != valid_blocks);
  543. #endif
  544. /* check segment usage, and check boundary of a given segment number */
  545. f2fs_bug_on(sbi, GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
  546. || segno > TOTAL_SEGS(sbi) - 1);
  547. }
  548. static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
  549. unsigned int start)
  550. {
  551. struct sit_info *sit_i = SIT_I(sbi);
  552. unsigned int offset = SIT_BLOCK_OFFSET(start);
  553. block_t blk_addr = sit_i->sit_base_addr + offset;
  554. check_seg_range(sbi, start);
  555. /* calculate sit block address */
  556. if (f2fs_test_bit(offset, sit_i->sit_bitmap))
  557. blk_addr += sit_i->sit_blocks;
  558. return blk_addr;
  559. }
  560. static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
  561. pgoff_t block_addr)
  562. {
  563. struct sit_info *sit_i = SIT_I(sbi);
  564. block_addr -= sit_i->sit_base_addr;
  565. if (block_addr < sit_i->sit_blocks)
  566. block_addr += sit_i->sit_blocks;
  567. else
  568. block_addr -= sit_i->sit_blocks;
  569. return block_addr + sit_i->sit_base_addr;
  570. }
  571. static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
  572. {
  573. unsigned int block_off = SIT_BLOCK_OFFSET(start);
  574. f2fs_change_bit(block_off, sit_i->sit_bitmap);
  575. }
  576. static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
  577. {
  578. struct sit_info *sit_i = SIT_I(sbi);
  579. return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec -
  580. sit_i->mounted_time;
  581. }
  582. static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
  583. unsigned int ofs_in_node, unsigned char version)
  584. {
  585. sum->nid = cpu_to_le32(nid);
  586. sum->ofs_in_node = cpu_to_le16(ofs_in_node);
  587. sum->version = version;
  588. }
  589. static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
  590. {
  591. return __start_cp_addr(sbi) +
  592. le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
  593. }
  594. static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
  595. {
  596. return __start_cp_addr(sbi) +
  597. le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
  598. - (base + 1) + type;
  599. }
  600. static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
  601. {
  602. if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
  603. return true;
  604. return false;
  605. }
  606. /*
  607. * It is very important to gather dirty pages and write at once, so that we can
  608. * submit a big bio without interfering other data writes.
  609. * By default, 512 pages for directory data,
  610. * 512 pages (2MB) * 3 for three types of nodes, and
  611. * max_bio_blocks for meta are set.
  612. */
  613. static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
  614. {
  615. if (sbi->sb->s_bdi->wb.dirty_exceeded)
  616. return 0;
  617. if (type == DATA)
  618. return sbi->blocks_per_seg;
  619. else if (type == NODE)
  620. return 8 * sbi->blocks_per_seg;
  621. else if (type == META)
  622. return 8 * BIO_MAX_PAGES;
  623. else
  624. return 0;
  625. }
  626. /*
  627. * When writing pages, it'd better align nr_to_write for segment size.
  628. */
  629. static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
  630. struct writeback_control *wbc)
  631. {
  632. long nr_to_write, desired;
  633. if (wbc->sync_mode != WB_SYNC_NONE)
  634. return 0;
  635. nr_to_write = wbc->nr_to_write;
  636. desired = BIO_MAX_PAGES;
  637. if (type == NODE)
  638. desired <<= 1;
  639. wbc->nr_to_write = desired;
  640. return desired - nr_to_write;
  641. }