node.h 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415
  1. /*
  2. * fs/f2fs/node.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. /* start node id of a node block dedicated to the given node id */
  12. #define START_NID(nid) ((nid / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
  13. /* node block offset on the NAT area dedicated to the given start node id */
  14. #define NAT_BLOCK_OFFSET(start_nid) (start_nid / NAT_ENTRY_PER_BLOCK)
  15. /* # of pages to perform readahead before building free nids */
  16. #define FREE_NID_PAGES 4
  17. /* maximum readahead size for node during getting data blocks */
  18. #define MAX_RA_NODE 128
  19. /* control the memory footprint threshold (10MB per 1GB ram) */
  20. #define DEF_RAM_THRESHOLD 10
  21. /* vector size for gang look-up from nat cache that consists of radix tree */
  22. #define NATVEC_SIZE 64
  23. #define SETVEC_SIZE 32
  24. /* return value for read_node_page */
  25. #define LOCKED_PAGE 1
  26. /* For flag in struct node_info */
  27. enum {
  28. IS_CHECKPOINTED, /* is it checkpointed before? */
  29. HAS_FSYNCED_INODE, /* is the inode fsynced before? */
  30. HAS_LAST_FSYNC, /* has the latest node fsync mark? */
  31. IS_DIRTY, /* this nat entry is dirty? */
  32. };
  33. /*
  34. * For node information
  35. */
  36. struct node_info {
  37. nid_t nid; /* node id */
  38. nid_t ino; /* inode number of the node's owner */
  39. block_t blk_addr; /* block address of the node */
  40. unsigned char version; /* version of the node */
  41. unsigned char flag; /* for node information bits */
  42. };
  43. struct nat_entry {
  44. struct list_head list; /* for clean or dirty nat list */
  45. struct node_info ni; /* in-memory node information */
  46. };
  47. #define nat_get_nid(nat) (nat->ni.nid)
  48. #define nat_set_nid(nat, n) (nat->ni.nid = n)
  49. #define nat_get_blkaddr(nat) (nat->ni.blk_addr)
  50. #define nat_set_blkaddr(nat, b) (nat->ni.blk_addr = b)
  51. #define nat_get_ino(nat) (nat->ni.ino)
  52. #define nat_set_ino(nat, i) (nat->ni.ino = i)
  53. #define nat_get_version(nat) (nat->ni.version)
  54. #define nat_set_version(nat, v) (nat->ni.version = v)
  55. #define inc_node_version(version) (++version)
  56. static inline void copy_node_info(struct node_info *dst,
  57. struct node_info *src)
  58. {
  59. dst->nid = src->nid;
  60. dst->ino = src->ino;
  61. dst->blk_addr = src->blk_addr;
  62. dst->version = src->version;
  63. /* should not copy flag here */
  64. }
  65. static inline void set_nat_flag(struct nat_entry *ne,
  66. unsigned int type, bool set)
  67. {
  68. unsigned char mask = 0x01 << type;
  69. if (set)
  70. ne->ni.flag |= mask;
  71. else
  72. ne->ni.flag &= ~mask;
  73. }
  74. static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
  75. {
  76. unsigned char mask = 0x01 << type;
  77. return ne->ni.flag & mask;
  78. }
  79. static inline void nat_reset_flag(struct nat_entry *ne)
  80. {
  81. /* these states can be set only after checkpoint was done */
  82. set_nat_flag(ne, IS_CHECKPOINTED, true);
  83. set_nat_flag(ne, HAS_FSYNCED_INODE, false);
  84. set_nat_flag(ne, HAS_LAST_FSYNC, true);
  85. }
  86. static inline void node_info_from_raw_nat(struct node_info *ni,
  87. struct f2fs_nat_entry *raw_ne)
  88. {
  89. ni->ino = le32_to_cpu(raw_ne->ino);
  90. ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
  91. ni->version = raw_ne->version;
  92. }
  93. static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
  94. struct node_info *ni)
  95. {
  96. raw_ne->ino = cpu_to_le32(ni->ino);
  97. raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
  98. raw_ne->version = ni->version;
  99. }
  100. enum mem_type {
  101. FREE_NIDS, /* indicates the free nid list */
  102. NAT_ENTRIES, /* indicates the cached nat entry */
  103. DIRTY_DENTS, /* indicates dirty dentry pages */
  104. INO_ENTRIES, /* indicates inode entries */
  105. BASE_CHECK, /* check kernel status */
  106. };
  107. struct nat_entry_set {
  108. struct list_head set_list; /* link with other nat sets */
  109. struct list_head entry_list; /* link with dirty nat entries */
  110. nid_t set; /* set number*/
  111. unsigned int entry_cnt; /* the # of nat entries in set */
  112. };
  113. /*
  114. * For free nid mangement
  115. */
  116. enum nid_state {
  117. NID_NEW, /* newly added to free nid list */
  118. NID_ALLOC /* it is allocated */
  119. };
  120. struct free_nid {
  121. struct list_head list; /* for free node id list */
  122. nid_t nid; /* node id */
  123. int state; /* in use or not: NID_NEW or NID_ALLOC */
  124. };
  125. static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
  126. {
  127. struct f2fs_nm_info *nm_i = NM_I(sbi);
  128. struct free_nid *fnid;
  129. spin_lock(&nm_i->free_nid_list_lock);
  130. if (nm_i->fcnt <= 0) {
  131. spin_unlock(&nm_i->free_nid_list_lock);
  132. return;
  133. }
  134. fnid = list_entry(nm_i->free_nid_list.next, struct free_nid, list);
  135. *nid = fnid->nid;
  136. spin_unlock(&nm_i->free_nid_list_lock);
  137. }
  138. /*
  139. * inline functions
  140. */
  141. static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
  142. {
  143. struct f2fs_nm_info *nm_i = NM_I(sbi);
  144. memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
  145. }
  146. static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
  147. {
  148. struct f2fs_nm_info *nm_i = NM_I(sbi);
  149. pgoff_t block_off;
  150. pgoff_t block_addr;
  151. int seg_off;
  152. block_off = NAT_BLOCK_OFFSET(start);
  153. seg_off = block_off >> sbi->log_blocks_per_seg;
  154. block_addr = (pgoff_t)(nm_i->nat_blkaddr +
  155. (seg_off << sbi->log_blocks_per_seg << 1) +
  156. (block_off & ((1 << sbi->log_blocks_per_seg) - 1)));
  157. if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
  158. block_addr += sbi->blocks_per_seg;
  159. return block_addr;
  160. }
  161. static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
  162. pgoff_t block_addr)
  163. {
  164. struct f2fs_nm_info *nm_i = NM_I(sbi);
  165. block_addr -= nm_i->nat_blkaddr;
  166. if ((block_addr >> sbi->log_blocks_per_seg) % 2)
  167. block_addr -= sbi->blocks_per_seg;
  168. else
  169. block_addr += sbi->blocks_per_seg;
  170. return block_addr + nm_i->nat_blkaddr;
  171. }
  172. static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
  173. {
  174. unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
  175. f2fs_change_bit(block_off, nm_i->nat_bitmap);
  176. }
  177. static inline void fill_node_footer(struct page *page, nid_t nid,
  178. nid_t ino, unsigned int ofs, bool reset)
  179. {
  180. struct f2fs_node *rn = F2FS_NODE(page);
  181. unsigned int old_flag = 0;
  182. if (reset)
  183. memset(rn, 0, sizeof(*rn));
  184. else
  185. old_flag = le32_to_cpu(rn->footer.flag);
  186. rn->footer.nid = cpu_to_le32(nid);
  187. rn->footer.ino = cpu_to_le32(ino);
  188. /* should remain old flag bits such as COLD_BIT_SHIFT */
  189. rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
  190. (old_flag & OFFSET_BIT_MASK));
  191. }
  192. static inline void copy_node_footer(struct page *dst, struct page *src)
  193. {
  194. struct f2fs_node *src_rn = F2FS_NODE(src);
  195. struct f2fs_node *dst_rn = F2FS_NODE(dst);
  196. memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
  197. }
  198. static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
  199. {
  200. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  201. struct f2fs_node *rn = F2FS_NODE(page);
  202. rn->footer.cp_ver = ckpt->checkpoint_ver;
  203. rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
  204. }
  205. static inline nid_t ino_of_node(struct page *node_page)
  206. {
  207. struct f2fs_node *rn = F2FS_NODE(node_page);
  208. return le32_to_cpu(rn->footer.ino);
  209. }
  210. static inline nid_t nid_of_node(struct page *node_page)
  211. {
  212. struct f2fs_node *rn = F2FS_NODE(node_page);
  213. return le32_to_cpu(rn->footer.nid);
  214. }
  215. static inline unsigned int ofs_of_node(struct page *node_page)
  216. {
  217. struct f2fs_node *rn = F2FS_NODE(node_page);
  218. unsigned flag = le32_to_cpu(rn->footer.flag);
  219. return flag >> OFFSET_BIT_SHIFT;
  220. }
  221. static inline unsigned long long cpver_of_node(struct page *node_page)
  222. {
  223. struct f2fs_node *rn = F2FS_NODE(node_page);
  224. return le64_to_cpu(rn->footer.cp_ver);
  225. }
  226. static inline block_t next_blkaddr_of_node(struct page *node_page)
  227. {
  228. struct f2fs_node *rn = F2FS_NODE(node_page);
  229. return le32_to_cpu(rn->footer.next_blkaddr);
  230. }
  231. /*
  232. * f2fs assigns the following node offsets described as (num).
  233. * N = NIDS_PER_BLOCK
  234. *
  235. * Inode block (0)
  236. * |- direct node (1)
  237. * |- direct node (2)
  238. * |- indirect node (3)
  239. * | `- direct node (4 => 4 + N - 1)
  240. * |- indirect node (4 + N)
  241. * | `- direct node (5 + N => 5 + 2N - 1)
  242. * `- double indirect node (5 + 2N)
  243. * `- indirect node (6 + 2N)
  244. * `- direct node
  245. * ......
  246. * `- indirect node ((6 + 2N) + x(N + 1))
  247. * `- direct node
  248. * ......
  249. * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
  250. * `- direct node
  251. */
  252. static inline bool IS_DNODE(struct page *node_page)
  253. {
  254. unsigned int ofs = ofs_of_node(node_page);
  255. if (f2fs_has_xattr_block(ofs))
  256. return false;
  257. if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
  258. ofs == 5 + 2 * NIDS_PER_BLOCK)
  259. return false;
  260. if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
  261. ofs -= 6 + 2 * NIDS_PER_BLOCK;
  262. if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
  263. return false;
  264. }
  265. return true;
  266. }
  267. static inline void set_nid(struct page *p, int off, nid_t nid, bool i)
  268. {
  269. struct f2fs_node *rn = F2FS_NODE(p);
  270. f2fs_wait_on_page_writeback(p, NODE);
  271. if (i)
  272. rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
  273. else
  274. rn->in.nid[off] = cpu_to_le32(nid);
  275. set_page_dirty(p);
  276. }
  277. static inline nid_t get_nid(struct page *p, int off, bool i)
  278. {
  279. struct f2fs_node *rn = F2FS_NODE(p);
  280. if (i)
  281. return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
  282. return le32_to_cpu(rn->in.nid[off]);
  283. }
  284. /*
  285. * Coldness identification:
  286. * - Mark cold files in f2fs_inode_info
  287. * - Mark cold node blocks in their node footer
  288. * - Mark cold data pages in page cache
  289. */
  290. static inline int is_file(struct inode *inode, int type)
  291. {
  292. return F2FS_I(inode)->i_advise & type;
  293. }
  294. static inline void set_file(struct inode *inode, int type)
  295. {
  296. F2FS_I(inode)->i_advise |= type;
  297. }
  298. static inline void clear_file(struct inode *inode, int type)
  299. {
  300. F2FS_I(inode)->i_advise &= ~type;
  301. }
  302. #define file_is_cold(inode) is_file(inode, FADVISE_COLD_BIT)
  303. #define file_wrong_pino(inode) is_file(inode, FADVISE_LOST_PINO_BIT)
  304. #define file_set_cold(inode) set_file(inode, FADVISE_COLD_BIT)
  305. #define file_lost_pino(inode) set_file(inode, FADVISE_LOST_PINO_BIT)
  306. #define file_clear_cold(inode) clear_file(inode, FADVISE_COLD_BIT)
  307. #define file_got_pino(inode) clear_file(inode, FADVISE_LOST_PINO_BIT)
  308. static inline int is_cold_data(struct page *page)
  309. {
  310. return PageChecked(page);
  311. }
  312. static inline void set_cold_data(struct page *page)
  313. {
  314. SetPageChecked(page);
  315. }
  316. static inline void clear_cold_data(struct page *page)
  317. {
  318. ClearPageChecked(page);
  319. }
  320. static inline int is_node(struct page *page, int type)
  321. {
  322. struct f2fs_node *rn = F2FS_NODE(page);
  323. return le32_to_cpu(rn->footer.flag) & (1 << type);
  324. }
  325. #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
  326. #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
  327. #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
  328. static inline void set_cold_node(struct inode *inode, struct page *page)
  329. {
  330. struct f2fs_node *rn = F2FS_NODE(page);
  331. unsigned int flag = le32_to_cpu(rn->footer.flag);
  332. if (S_ISDIR(inode->i_mode))
  333. flag &= ~(0x1 << COLD_BIT_SHIFT);
  334. else
  335. flag |= (0x1 << COLD_BIT_SHIFT);
  336. rn->footer.flag = cpu_to_le32(flag);
  337. }
  338. static inline void set_mark(struct page *page, int mark, int type)
  339. {
  340. struct f2fs_node *rn = F2FS_NODE(page);
  341. unsigned int flag = le32_to_cpu(rn->footer.flag);
  342. if (mark)
  343. flag |= (0x1 << type);
  344. else
  345. flag &= ~(0x1 << type);
  346. rn->footer.flag = cpu_to_le32(flag);
  347. }
  348. #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
  349. #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)