segment.h 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763
  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 F2FS_MIN_SEGMENTS 9 /* SB + 2 (CP + SIT + NAT) + SSA + MAIN */
  18. /* L: Logical segment # in volume, R: Relative segment # in main area */
  19. #define GET_L2R_SEGNO(free_i, segno) (segno - free_i->start_segno)
  20. #define GET_R2L_SEGNO(free_i, segno) (segno + free_i->start_segno)
  21. #define IS_DATASEG(t) (t <= CURSEG_COLD_DATA)
  22. #define IS_NODESEG(t) (t >= CURSEG_HOT_NODE)
  23. #define IS_CURSEG(sbi, seg) \
  24. ((seg == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno) || \
  25. (seg == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno) || \
  26. (seg == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno) || \
  27. (seg == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) || \
  28. (seg == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) || \
  29. (seg == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno))
  30. #define IS_CURSEC(sbi, secno) \
  31. ((secno == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno / \
  32. sbi->segs_per_sec) || \
  33. (secno == CURSEG_I(sbi, CURSEG_WARM_DATA)->segno / \
  34. sbi->segs_per_sec) || \
  35. (secno == CURSEG_I(sbi, CURSEG_COLD_DATA)->segno / \
  36. sbi->segs_per_sec) || \
  37. (secno == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno / \
  38. sbi->segs_per_sec) || \
  39. (secno == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno / \
  40. sbi->segs_per_sec) || \
  41. (secno == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno / \
  42. sbi->segs_per_sec)) \
  43. #define MAIN_BLKADDR(sbi) \
  44. (SM_I(sbi) ? SM_I(sbi)->main_blkaddr : \
  45. le32_to_cpu(F2FS_RAW_SUPER(sbi)->main_blkaddr))
  46. #define SEG0_BLKADDR(sbi) \
  47. (SM_I(sbi) ? SM_I(sbi)->seg0_blkaddr : \
  48. le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment0_blkaddr))
  49. #define MAIN_SEGS(sbi) (SM_I(sbi)->main_segments)
  50. #define MAIN_SECS(sbi) (sbi->total_sections)
  51. #define TOTAL_SEGS(sbi) \
  52. (SM_I(sbi) ? SM_I(sbi)->segment_count : \
  53. le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count))
  54. #define TOTAL_BLKS(sbi) (TOTAL_SEGS(sbi) << sbi->log_blocks_per_seg)
  55. #define MAX_BLKADDR(sbi) (SEG0_BLKADDR(sbi) + TOTAL_BLKS(sbi))
  56. #define SEGMENT_SIZE(sbi) (1ULL << (sbi->log_blocksize + \
  57. sbi->log_blocks_per_seg))
  58. #define START_BLOCK(sbi, segno) (SEG0_BLKADDR(sbi) + \
  59. (GET_R2L_SEGNO(FREE_I(sbi), segno) << sbi->log_blocks_per_seg))
  60. #define NEXT_FREE_BLKADDR(sbi, curseg) \
  61. (START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff)
  62. #define GET_SEGOFF_FROM_SEG0(sbi, blk_addr) ((blk_addr) - SEG0_BLKADDR(sbi))
  63. #define GET_SEGNO_FROM_SEG0(sbi, blk_addr) \
  64. (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) >> sbi->log_blocks_per_seg)
  65. #define GET_BLKOFF_FROM_SEG0(sbi, blk_addr) \
  66. (GET_SEGOFF_FROM_SEG0(sbi, blk_addr) & (sbi->blocks_per_seg - 1))
  67. #define GET_SEGNO(sbi, blk_addr) \
  68. ((!is_valid_data_blkaddr(sbi, blk_addr)) ? \
  69. NULL_SEGNO : GET_L2R_SEGNO(FREE_I(sbi), \
  70. GET_SEGNO_FROM_SEG0(sbi, blk_addr)))
  71. #define GET_SECNO(sbi, segno) \
  72. ((segno) / sbi->segs_per_sec)
  73. #define GET_ZONENO_FROM_SEGNO(sbi, segno) \
  74. ((segno / sbi->segs_per_sec) / sbi->secs_per_zone)
  75. #define GET_SUM_BLOCK(sbi, segno) \
  76. ((sbi->sm_info->ssa_blkaddr) + segno)
  77. #define GET_SUM_TYPE(footer) ((footer)->entry_type)
  78. #define SET_SUM_TYPE(footer, type) ((footer)->entry_type = type)
  79. #define SIT_ENTRY_OFFSET(sit_i, segno) \
  80. (segno % sit_i->sents_per_block)
  81. #define SIT_BLOCK_OFFSET(segno) \
  82. (segno / SIT_ENTRY_PER_BLOCK)
  83. #define START_SEGNO(segno) \
  84. (SIT_BLOCK_OFFSET(segno) * SIT_ENTRY_PER_BLOCK)
  85. #define SIT_BLK_CNT(sbi) \
  86. ((MAIN_SEGS(sbi) + SIT_ENTRY_PER_BLOCK - 1) / SIT_ENTRY_PER_BLOCK)
  87. #define f2fs_bitmap_size(nr) \
  88. (BITS_TO_LONGS(nr) * sizeof(unsigned long))
  89. #define SECTOR_FROM_BLOCK(blk_addr) \
  90. (((sector_t)blk_addr) << F2FS_LOG_SECTORS_PER_BLOCK)
  91. #define SECTOR_TO_BLOCK(sectors) \
  92. (sectors >> F2FS_LOG_SECTORS_PER_BLOCK)
  93. #define MAX_BIO_BLOCKS(sbi) \
  94. ((int)min((int)max_hw_blocks(sbi), BIO_MAX_PAGES))
  95. /*
  96. * indicate a block allocation direction: RIGHT and LEFT.
  97. * RIGHT means allocating new sections towards the end of volume.
  98. * LEFT means the opposite direction.
  99. */
  100. enum {
  101. ALLOC_RIGHT = 0,
  102. ALLOC_LEFT
  103. };
  104. /*
  105. * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
  106. * LFS writes data sequentially with cleaning operations.
  107. * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
  108. */
  109. enum {
  110. LFS = 0,
  111. SSR
  112. };
  113. /*
  114. * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
  115. * GC_CB is based on cost-benefit algorithm.
  116. * GC_GREEDY is based on greedy algorithm.
  117. */
  118. enum {
  119. GC_CB = 0,
  120. GC_GREEDY
  121. };
  122. /*
  123. * BG_GC means the background cleaning job.
  124. * FG_GC means the on-demand cleaning job.
  125. * FORCE_FG_GC means on-demand cleaning job in background.
  126. */
  127. enum {
  128. BG_GC = 0,
  129. FG_GC,
  130. FORCE_FG_GC,
  131. };
  132. /* for a function parameter to select a victim segment */
  133. struct victim_sel_policy {
  134. int alloc_mode; /* LFS or SSR */
  135. int gc_mode; /* GC_CB or GC_GREEDY */
  136. unsigned long *dirty_segmap; /* dirty segment bitmap */
  137. unsigned int max_search; /* maximum # of segments to search */
  138. unsigned int offset; /* last scanned bitmap offset */
  139. unsigned int ofs_unit; /* bitmap search unit */
  140. unsigned int min_cost; /* minimum cost */
  141. unsigned int min_segno; /* segment # having min. cost */
  142. };
  143. struct seg_entry {
  144. unsigned short valid_blocks; /* # of valid blocks */
  145. unsigned char *cur_valid_map; /* validity bitmap of blocks */
  146. /*
  147. * # of valid blocks and the validity bitmap stored in the the last
  148. * checkpoint pack. This information is used by the SSR mode.
  149. */
  150. unsigned short ckpt_valid_blocks;
  151. unsigned char *ckpt_valid_map;
  152. unsigned char *discard_map;
  153. unsigned char type; /* segment type like CURSEG_XXX_TYPE */
  154. unsigned long long mtime; /* modification time of the segment */
  155. };
  156. struct sec_entry {
  157. unsigned int valid_blocks; /* # of valid blocks in a section */
  158. };
  159. struct segment_allocation {
  160. void (*allocate_segment)(struct f2fs_sb_info *, int, bool);
  161. };
  162. /*
  163. * this value is set in page as a private data which indicate that
  164. * the page is atomically written, and it is in inmem_pages list.
  165. */
  166. #define ATOMIC_WRITTEN_PAGE 0x0000ffff
  167. #define IS_ATOMIC_WRITTEN_PAGE(page) \
  168. (page_private(page) == (unsigned long)ATOMIC_WRITTEN_PAGE)
  169. struct inmem_pages {
  170. struct list_head list;
  171. struct page *page;
  172. };
  173. struct sit_info {
  174. const struct segment_allocation *s_ops;
  175. block_t sit_base_addr; /* start block address of SIT area */
  176. block_t sit_blocks; /* # of blocks used by SIT area */
  177. block_t written_valid_blocks; /* # of valid blocks in main area */
  178. char *sit_bitmap; /* SIT bitmap pointer */
  179. unsigned int bitmap_size; /* SIT bitmap size */
  180. unsigned long *tmp_map; /* bitmap for temporal use */
  181. unsigned long *dirty_sentries_bitmap; /* bitmap for dirty sentries */
  182. unsigned int dirty_sentries; /* # of dirty sentries */
  183. unsigned int sents_per_block; /* # of SIT entries per block */
  184. struct mutex sentry_lock; /* to protect SIT cache */
  185. struct seg_entry *sentries; /* SIT segment-level cache */
  186. struct sec_entry *sec_entries; /* SIT section-level cache */
  187. /* for cost-benefit algorithm in cleaning procedure */
  188. unsigned long long elapsed_time; /* elapsed time after mount */
  189. unsigned long long mounted_time; /* mount time */
  190. unsigned long long min_mtime; /* min. modification time */
  191. unsigned long long max_mtime; /* max. modification time */
  192. };
  193. struct free_segmap_info {
  194. unsigned int start_segno; /* start segment number logically */
  195. unsigned int free_segments; /* # of free segments */
  196. unsigned int free_sections; /* # of free sections */
  197. spinlock_t segmap_lock; /* free segmap lock */
  198. unsigned long *free_segmap; /* free segment bitmap */
  199. unsigned long *free_secmap; /* free section bitmap */
  200. };
  201. /* Notice: The order of dirty type is same with CURSEG_XXX in f2fs.h */
  202. enum dirty_type {
  203. DIRTY_HOT_DATA, /* dirty segments assigned as hot data logs */
  204. DIRTY_WARM_DATA, /* dirty segments assigned as warm data logs */
  205. DIRTY_COLD_DATA, /* dirty segments assigned as cold data logs */
  206. DIRTY_HOT_NODE, /* dirty segments assigned as hot node logs */
  207. DIRTY_WARM_NODE, /* dirty segments assigned as warm node logs */
  208. DIRTY_COLD_NODE, /* dirty segments assigned as cold node logs */
  209. DIRTY, /* to count # of dirty segments */
  210. PRE, /* to count # of entirely obsolete segments */
  211. NR_DIRTY_TYPE
  212. };
  213. struct dirty_seglist_info {
  214. const struct victim_selection *v_ops; /* victim selction operation */
  215. unsigned long *dirty_segmap[NR_DIRTY_TYPE];
  216. struct mutex seglist_lock; /* lock for segment bitmaps */
  217. int nr_dirty[NR_DIRTY_TYPE]; /* # of dirty segments */
  218. unsigned long *victim_secmap; /* background GC victims */
  219. };
  220. /* victim selection function for cleaning and SSR */
  221. struct victim_selection {
  222. int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
  223. int, int, char);
  224. };
  225. /* for active log information */
  226. struct curseg_info {
  227. struct mutex curseg_mutex; /* lock for consistency */
  228. struct f2fs_summary_block *sum_blk; /* cached summary block */
  229. unsigned char alloc_type; /* current allocation type */
  230. unsigned int segno; /* current segment number */
  231. unsigned short next_blkoff; /* next block offset to write */
  232. unsigned int zone; /* current zone number */
  233. unsigned int next_segno; /* preallocated segment */
  234. };
  235. struct sit_entry_set {
  236. struct list_head set_list; /* link with all sit sets */
  237. unsigned int start_segno; /* start segno of sits in set */
  238. unsigned int entry_cnt; /* the # of sit entries in set */
  239. };
  240. /*
  241. * inline functions
  242. */
  243. static inline struct curseg_info *CURSEG_I(struct f2fs_sb_info *sbi, int type)
  244. {
  245. return (struct curseg_info *)(SM_I(sbi)->curseg_array + type);
  246. }
  247. static inline struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
  248. unsigned int segno)
  249. {
  250. struct sit_info *sit_i = SIT_I(sbi);
  251. return &sit_i->sentries[segno];
  252. }
  253. static inline struct sec_entry *get_sec_entry(struct f2fs_sb_info *sbi,
  254. unsigned int segno)
  255. {
  256. struct sit_info *sit_i = SIT_I(sbi);
  257. return &sit_i->sec_entries[GET_SECNO(sbi, segno)];
  258. }
  259. static inline unsigned int get_valid_blocks(struct f2fs_sb_info *sbi,
  260. unsigned int segno, int section)
  261. {
  262. /*
  263. * In order to get # of valid blocks in a section instantly from many
  264. * segments, f2fs manages two counting structures separately.
  265. */
  266. if (section > 1)
  267. return get_sec_entry(sbi, segno)->valid_blocks;
  268. else
  269. return get_seg_entry(sbi, segno)->valid_blocks;
  270. }
  271. static inline void seg_info_from_raw_sit(struct seg_entry *se,
  272. struct f2fs_sit_entry *rs)
  273. {
  274. se->valid_blocks = GET_SIT_VBLOCKS(rs);
  275. se->ckpt_valid_blocks = GET_SIT_VBLOCKS(rs);
  276. memcpy(se->cur_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  277. memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  278. se->type = GET_SIT_TYPE(rs);
  279. se->mtime = le64_to_cpu(rs->mtime);
  280. }
  281. static inline void seg_info_to_raw_sit(struct seg_entry *se,
  282. struct f2fs_sit_entry *rs)
  283. {
  284. unsigned short raw_vblocks = (se->type << SIT_VBLOCKS_SHIFT) |
  285. se->valid_blocks;
  286. rs->vblocks = cpu_to_le16(raw_vblocks);
  287. memcpy(rs->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
  288. memcpy(se->ckpt_valid_map, rs->valid_map, SIT_VBLOCK_MAP_SIZE);
  289. se->ckpt_valid_blocks = se->valid_blocks;
  290. rs->mtime = cpu_to_le64(se->mtime);
  291. }
  292. static inline unsigned int find_next_inuse(struct free_segmap_info *free_i,
  293. unsigned int max, unsigned int segno)
  294. {
  295. unsigned int ret;
  296. spin_lock(&free_i->segmap_lock);
  297. ret = find_next_bit(free_i->free_segmap, max, segno);
  298. spin_unlock(&free_i->segmap_lock);
  299. return ret;
  300. }
  301. static inline void __set_free(struct f2fs_sb_info *sbi, unsigned int segno)
  302. {
  303. struct free_segmap_info *free_i = FREE_I(sbi);
  304. unsigned int secno = segno / sbi->segs_per_sec;
  305. unsigned int start_segno = secno * sbi->segs_per_sec;
  306. unsigned int next;
  307. spin_lock(&free_i->segmap_lock);
  308. clear_bit(segno, free_i->free_segmap);
  309. free_i->free_segments++;
  310. next = find_next_bit(free_i->free_segmap,
  311. start_segno + sbi->segs_per_sec, start_segno);
  312. if (next >= start_segno + sbi->segs_per_sec) {
  313. clear_bit(secno, free_i->free_secmap);
  314. free_i->free_sections++;
  315. }
  316. spin_unlock(&free_i->segmap_lock);
  317. }
  318. static inline void __set_inuse(struct f2fs_sb_info *sbi,
  319. unsigned int segno)
  320. {
  321. struct free_segmap_info *free_i = FREE_I(sbi);
  322. unsigned int secno = segno / sbi->segs_per_sec;
  323. set_bit(segno, free_i->free_segmap);
  324. free_i->free_segments--;
  325. if (!test_and_set_bit(secno, free_i->free_secmap))
  326. free_i->free_sections--;
  327. }
  328. static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
  329. unsigned int segno)
  330. {
  331. struct free_segmap_info *free_i = FREE_I(sbi);
  332. unsigned int secno = segno / sbi->segs_per_sec;
  333. unsigned int start_segno = secno * sbi->segs_per_sec;
  334. unsigned int next;
  335. spin_lock(&free_i->segmap_lock);
  336. if (test_and_clear_bit(segno, free_i->free_segmap)) {
  337. free_i->free_segments++;
  338. if (IS_CURSEC(sbi, secno))
  339. goto skip_free;
  340. next = find_next_bit(free_i->free_segmap,
  341. start_segno + sbi->segs_per_sec, start_segno);
  342. if (next >= start_segno + sbi->segs_per_sec) {
  343. if (test_and_clear_bit(secno, free_i->free_secmap))
  344. free_i->free_sections++;
  345. }
  346. }
  347. skip_free:
  348. spin_unlock(&free_i->segmap_lock);
  349. }
  350. static inline void __set_test_and_inuse(struct f2fs_sb_info *sbi,
  351. unsigned int segno)
  352. {
  353. struct free_segmap_info *free_i = FREE_I(sbi);
  354. unsigned int secno = segno / sbi->segs_per_sec;
  355. spin_lock(&free_i->segmap_lock);
  356. if (!test_and_set_bit(segno, free_i->free_segmap)) {
  357. free_i->free_segments--;
  358. if (!test_and_set_bit(secno, free_i->free_secmap))
  359. free_i->free_sections--;
  360. }
  361. spin_unlock(&free_i->segmap_lock);
  362. }
  363. static inline void get_sit_bitmap(struct f2fs_sb_info *sbi,
  364. void *dst_addr)
  365. {
  366. struct sit_info *sit_i = SIT_I(sbi);
  367. memcpy(dst_addr, sit_i->sit_bitmap, sit_i->bitmap_size);
  368. }
  369. static inline block_t written_block_count(struct f2fs_sb_info *sbi)
  370. {
  371. return SIT_I(sbi)->written_valid_blocks;
  372. }
  373. static inline unsigned int free_segments(struct f2fs_sb_info *sbi)
  374. {
  375. return FREE_I(sbi)->free_segments;
  376. }
  377. static inline int reserved_segments(struct f2fs_sb_info *sbi)
  378. {
  379. return SM_I(sbi)->reserved_segments;
  380. }
  381. static inline unsigned int free_sections(struct f2fs_sb_info *sbi)
  382. {
  383. return FREE_I(sbi)->free_sections;
  384. }
  385. static inline unsigned int prefree_segments(struct f2fs_sb_info *sbi)
  386. {
  387. return DIRTY_I(sbi)->nr_dirty[PRE];
  388. }
  389. static inline unsigned int dirty_segments(struct f2fs_sb_info *sbi)
  390. {
  391. return DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_DATA] +
  392. DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_DATA] +
  393. DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_DATA] +
  394. DIRTY_I(sbi)->nr_dirty[DIRTY_HOT_NODE] +
  395. DIRTY_I(sbi)->nr_dirty[DIRTY_WARM_NODE] +
  396. DIRTY_I(sbi)->nr_dirty[DIRTY_COLD_NODE];
  397. }
  398. static inline int overprovision_segments(struct f2fs_sb_info *sbi)
  399. {
  400. return SM_I(sbi)->ovp_segments;
  401. }
  402. static inline int overprovision_sections(struct f2fs_sb_info *sbi)
  403. {
  404. return ((unsigned int) overprovision_segments(sbi)) / sbi->segs_per_sec;
  405. }
  406. static inline int reserved_sections(struct f2fs_sb_info *sbi)
  407. {
  408. return ((unsigned int) reserved_segments(sbi)) / sbi->segs_per_sec;
  409. }
  410. static inline bool need_SSR(struct f2fs_sb_info *sbi)
  411. {
  412. int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
  413. int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
  414. return free_sections(sbi) <= (node_secs + 2 * dent_secs +
  415. reserved_sections(sbi) + 1);
  416. }
  417. static inline bool has_not_enough_free_secs(struct f2fs_sb_info *sbi, int freed)
  418. {
  419. int node_secs = get_blocktype_secs(sbi, F2FS_DIRTY_NODES);
  420. int dent_secs = get_blocktype_secs(sbi, F2FS_DIRTY_DENTS);
  421. if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
  422. return false;
  423. return (free_sections(sbi) + freed) <= (node_secs + 2 * dent_secs +
  424. reserved_sections(sbi));
  425. }
  426. static inline bool excess_prefree_segs(struct f2fs_sb_info *sbi)
  427. {
  428. return prefree_segments(sbi) > SM_I(sbi)->rec_prefree_segments;
  429. }
  430. static inline int utilization(struct f2fs_sb_info *sbi)
  431. {
  432. return div_u64((u64)valid_user_blocks(sbi) * 100,
  433. sbi->user_block_count);
  434. }
  435. /*
  436. * Sometimes f2fs may be better to drop out-of-place update policy.
  437. * And, users can control the policy through sysfs entries.
  438. * There are five policies with triggering conditions as follows.
  439. * F2FS_IPU_FORCE - all the time,
  440. * F2FS_IPU_SSR - if SSR mode is activated,
  441. * F2FS_IPU_UTIL - if FS utilization is over threashold,
  442. * F2FS_IPU_SSR_UTIL - if SSR mode is activated and FS utilization is over
  443. * threashold,
  444. * F2FS_IPU_FSYNC - activated in fsync path only for high performance flash
  445. * storages. IPU will be triggered only if the # of dirty
  446. * pages over min_fsync_blocks.
  447. * F2FS_IPUT_DISABLE - disable IPU. (=default option)
  448. */
  449. #define DEF_MIN_IPU_UTIL 70
  450. #define DEF_MIN_FSYNC_BLOCKS 8
  451. enum {
  452. F2FS_IPU_FORCE,
  453. F2FS_IPU_SSR,
  454. F2FS_IPU_UTIL,
  455. F2FS_IPU_SSR_UTIL,
  456. F2FS_IPU_FSYNC,
  457. };
  458. static inline bool need_inplace_update(struct inode *inode)
  459. {
  460. struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
  461. unsigned int policy = SM_I(sbi)->ipu_policy;
  462. /* IPU can be done only for the user data */
  463. if (S_ISDIR(inode->i_mode) || f2fs_is_atomic_file(inode))
  464. return false;
  465. if (policy & (0x1 << F2FS_IPU_FORCE))
  466. return true;
  467. if (policy & (0x1 << F2FS_IPU_SSR) && need_SSR(sbi))
  468. return true;
  469. if (policy & (0x1 << F2FS_IPU_UTIL) &&
  470. utilization(sbi) > SM_I(sbi)->min_ipu_util)
  471. return true;
  472. if (policy & (0x1 << F2FS_IPU_SSR_UTIL) && need_SSR(sbi) &&
  473. utilization(sbi) > SM_I(sbi)->min_ipu_util)
  474. return true;
  475. /* this is only set during fdatasync */
  476. if (policy & (0x1 << F2FS_IPU_FSYNC) &&
  477. is_inode_flag_set(F2FS_I(inode), FI_NEED_IPU))
  478. return true;
  479. return false;
  480. }
  481. static inline unsigned int curseg_segno(struct f2fs_sb_info *sbi,
  482. int type)
  483. {
  484. struct curseg_info *curseg = CURSEG_I(sbi, type);
  485. return curseg->segno;
  486. }
  487. static inline unsigned char curseg_alloc_type(struct f2fs_sb_info *sbi,
  488. int type)
  489. {
  490. struct curseg_info *curseg = CURSEG_I(sbi, type);
  491. return curseg->alloc_type;
  492. }
  493. static inline unsigned short curseg_blkoff(struct f2fs_sb_info *sbi, int type)
  494. {
  495. struct curseg_info *curseg = CURSEG_I(sbi, type);
  496. return curseg->next_blkoff;
  497. }
  498. static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
  499. {
  500. f2fs_bug_on(sbi, segno > TOTAL_SEGS(sbi) - 1);
  501. }
  502. static inline void verify_block_addr(struct f2fs_io_info *fio, block_t blk_addr)
  503. {
  504. struct f2fs_sb_info *sbi = fio->sbi;
  505. if (__is_meta_io(fio))
  506. verify_blkaddr(sbi, blk_addr, META_GENERIC);
  507. else
  508. verify_blkaddr(sbi, blk_addr, DATA_GENERIC);
  509. }
  510. /*
  511. * Summary block is always treated as an invalid block
  512. */
  513. static inline int check_block_count(struct f2fs_sb_info *sbi,
  514. int segno, struct f2fs_sit_entry *raw_sit)
  515. {
  516. #ifdef CONFIG_F2FS_CHECK_FS
  517. bool is_valid = test_bit_le(0, raw_sit->valid_map) ? true : false;
  518. int valid_blocks = 0;
  519. int cur_pos = 0, next_pos;
  520. /* check bitmap with valid block count */
  521. do {
  522. if (is_valid) {
  523. next_pos = find_next_zero_bit_le(&raw_sit->valid_map,
  524. sbi->blocks_per_seg,
  525. cur_pos);
  526. valid_blocks += next_pos - cur_pos;
  527. } else
  528. next_pos = find_next_bit_le(&raw_sit->valid_map,
  529. sbi->blocks_per_seg,
  530. cur_pos);
  531. cur_pos = next_pos;
  532. is_valid = !is_valid;
  533. } while (cur_pos < sbi->blocks_per_seg);
  534. if (unlikely(GET_SIT_VBLOCKS(raw_sit) != valid_blocks)) {
  535. f2fs_msg(sbi->sb, KERN_ERR,
  536. "Mismatch valid blocks %d vs. %d",
  537. GET_SIT_VBLOCKS(raw_sit), valid_blocks);
  538. set_sbi_flag(sbi, SBI_NEED_FSCK);
  539. return -EINVAL;
  540. }
  541. #endif
  542. /* check segment usage, and check boundary of a given segment number */
  543. if (unlikely(GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg
  544. || segno > TOTAL_SEGS(sbi) - 1)) {
  545. f2fs_msg(sbi->sb, KERN_ERR,
  546. "Wrong valid blocks %d or segno %u",
  547. GET_SIT_VBLOCKS(raw_sit), segno);
  548. set_sbi_flag(sbi, SBI_NEED_FSCK);
  549. return -EINVAL;
  550. }
  551. return 0;
  552. }
  553. static inline pgoff_t current_sit_addr(struct f2fs_sb_info *sbi,
  554. unsigned int start)
  555. {
  556. struct sit_info *sit_i = SIT_I(sbi);
  557. unsigned int offset = SIT_BLOCK_OFFSET(start);
  558. block_t blk_addr = sit_i->sit_base_addr + offset;
  559. check_seg_range(sbi, start);
  560. /* calculate sit block address */
  561. if (f2fs_test_bit(offset, sit_i->sit_bitmap))
  562. blk_addr += sit_i->sit_blocks;
  563. return blk_addr;
  564. }
  565. static inline pgoff_t next_sit_addr(struct f2fs_sb_info *sbi,
  566. pgoff_t block_addr)
  567. {
  568. struct sit_info *sit_i = SIT_I(sbi);
  569. block_addr -= sit_i->sit_base_addr;
  570. if (block_addr < sit_i->sit_blocks)
  571. block_addr += sit_i->sit_blocks;
  572. else
  573. block_addr -= sit_i->sit_blocks;
  574. return block_addr + sit_i->sit_base_addr;
  575. }
  576. static inline void set_to_next_sit(struct sit_info *sit_i, unsigned int start)
  577. {
  578. unsigned int block_off = SIT_BLOCK_OFFSET(start);
  579. f2fs_change_bit(block_off, sit_i->sit_bitmap);
  580. }
  581. static inline unsigned long long get_mtime(struct f2fs_sb_info *sbi)
  582. {
  583. struct sit_info *sit_i = SIT_I(sbi);
  584. return sit_i->elapsed_time + CURRENT_TIME_SEC.tv_sec -
  585. sit_i->mounted_time;
  586. }
  587. static inline void set_summary(struct f2fs_summary *sum, nid_t nid,
  588. unsigned int ofs_in_node, unsigned char version)
  589. {
  590. sum->nid = cpu_to_le32(nid);
  591. sum->ofs_in_node = cpu_to_le16(ofs_in_node);
  592. sum->version = version;
  593. }
  594. static inline block_t start_sum_block(struct f2fs_sb_info *sbi)
  595. {
  596. return __start_cp_addr(sbi) +
  597. le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_start_sum);
  598. }
  599. static inline block_t sum_blk_addr(struct f2fs_sb_info *sbi, int base, int type)
  600. {
  601. return __start_cp_addr(sbi) +
  602. le32_to_cpu(F2FS_CKPT(sbi)->cp_pack_total_block_count)
  603. - (base + 1) + type;
  604. }
  605. static inline bool sec_usage_check(struct f2fs_sb_info *sbi, unsigned int secno)
  606. {
  607. if (IS_CURSEC(sbi, secno) || (sbi->cur_victim_sec == secno))
  608. return true;
  609. return false;
  610. }
  611. static inline unsigned int max_hw_blocks(struct f2fs_sb_info *sbi)
  612. {
  613. struct block_device *bdev = sbi->sb->s_bdev;
  614. struct request_queue *q = bdev_get_queue(bdev);
  615. return SECTOR_TO_BLOCK(queue_max_sectors(q));
  616. }
  617. /*
  618. * It is very important to gather dirty pages and write at once, so that we can
  619. * submit a big bio without interfering other data writes.
  620. * By default, 512 pages for directory data,
  621. * 512 pages (2MB) * 3 for three types of nodes, and
  622. * max_bio_blocks for meta are set.
  623. */
  624. static inline int nr_pages_to_skip(struct f2fs_sb_info *sbi, int type)
  625. {
  626. if (sbi->sb->s_bdi->wb.dirty_exceeded)
  627. return 0;
  628. if (type == DATA)
  629. return sbi->blocks_per_seg;
  630. else if (type == NODE)
  631. return 3 * sbi->blocks_per_seg;
  632. else if (type == META)
  633. return MAX_BIO_BLOCKS(sbi);
  634. else
  635. return 0;
  636. }
  637. /*
  638. * When writing pages, it'd better align nr_to_write for segment size.
  639. */
  640. static inline long nr_pages_to_write(struct f2fs_sb_info *sbi, int type,
  641. struct writeback_control *wbc)
  642. {
  643. long nr_to_write, desired;
  644. if (wbc->sync_mode != WB_SYNC_NONE)
  645. return 0;
  646. nr_to_write = wbc->nr_to_write;
  647. if (type == DATA)
  648. desired = 4096;
  649. else if (type == NODE)
  650. desired = 3 * max_hw_blocks(sbi);
  651. else
  652. desired = MAX_BIO_BLOCKS(sbi);
  653. wbc->nr_to_write = desired;
  654. return desired - nr_to_write;
  655. }