node.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455
  1. // SPDX-License-Identifier: GPL-2.0
  2. /*
  3. * fs/f2fs/node.h
  4. *
  5. * Copyright (c) 2012 Samsung Electronics Co., Ltd.
  6. * http://www.samsung.com/
  7. */
  8. /* start node id of a node block dedicated to the given node id */
  9. #define START_NID(nid) (((nid) / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
  10. /* node block offset on the NAT area dedicated to the given start node id */
  11. #define NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK)
  12. /* # of pages to perform synchronous readahead before building free nids */
  13. #define FREE_NID_PAGES 8
  14. #define MAX_FREE_NIDS (NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES)
  15. #define DEF_RA_NID_PAGES 0 /* # of nid pages to be readaheaded */
  16. /* maximum readahead size for node during getting data blocks */
  17. #define MAX_RA_NODE 128
  18. /* control the memory footprint threshold (10MB per 1GB ram) */
  19. #define DEF_RAM_THRESHOLD 1
  20. /* control dirty nats ratio threshold (default: 10% over max nid count) */
  21. #define DEF_DIRTY_NAT_RATIO_THRESHOLD 10
  22. /* control total # of nats */
  23. #define DEF_NAT_CACHE_THRESHOLD 100000
  24. /* vector size for gang look-up from nat cache that consists of radix tree */
  25. #define NATVEC_SIZE 64
  26. #define SETVEC_SIZE 32
  27. /* return value for read_node_page */
  28. #define LOCKED_PAGE 1
  29. /* For flag in struct node_info */
  30. enum {
  31. IS_CHECKPOINTED, /* is it checkpointed before? */
  32. HAS_FSYNCED_INODE, /* is the inode fsynced before? */
  33. HAS_LAST_FSYNC, /* has the latest node fsync mark? */
  34. IS_DIRTY, /* this nat entry is dirty? */
  35. IS_PREALLOC, /* nat entry is preallocated */
  36. };
  37. /*
  38. * For node information
  39. */
  40. struct node_info {
  41. nid_t nid; /* node id */
  42. nid_t ino; /* inode number of the node's owner */
  43. block_t blk_addr; /* block address of the node */
  44. unsigned char version; /* version of the node */
  45. unsigned char flag; /* for node information bits */
  46. };
  47. struct nat_entry {
  48. struct list_head list; /* for clean or dirty nat list */
  49. struct node_info ni; /* in-memory node information */
  50. };
  51. #define nat_get_nid(nat) ((nat)->ni.nid)
  52. #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
  53. #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
  54. #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
  55. #define nat_get_ino(nat) ((nat)->ni.ino)
  56. #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
  57. #define nat_get_version(nat) ((nat)->ni.version)
  58. #define nat_set_version(nat, v) ((nat)->ni.version = (v))
  59. #define inc_node_version(version) (++(version))
  60. static inline void copy_node_info(struct node_info *dst,
  61. struct node_info *src)
  62. {
  63. dst->nid = src->nid;
  64. dst->ino = src->ino;
  65. dst->blk_addr = src->blk_addr;
  66. dst->version = src->version;
  67. /* should not copy flag here */
  68. }
  69. static inline void set_nat_flag(struct nat_entry *ne,
  70. unsigned int type, bool set)
  71. {
  72. unsigned char mask = 0x01 << type;
  73. if (set)
  74. ne->ni.flag |= mask;
  75. else
  76. ne->ni.flag &= ~mask;
  77. }
  78. static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
  79. {
  80. unsigned char mask = 0x01 << type;
  81. return ne->ni.flag & mask;
  82. }
  83. static inline void nat_reset_flag(struct nat_entry *ne)
  84. {
  85. /* these states can be set only after checkpoint was done */
  86. set_nat_flag(ne, IS_CHECKPOINTED, true);
  87. set_nat_flag(ne, HAS_FSYNCED_INODE, false);
  88. set_nat_flag(ne, HAS_LAST_FSYNC, true);
  89. }
  90. static inline void node_info_from_raw_nat(struct node_info *ni,
  91. struct f2fs_nat_entry *raw_ne)
  92. {
  93. ni->ino = le32_to_cpu(raw_ne->ino);
  94. ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
  95. ni->version = raw_ne->version;
  96. }
  97. static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
  98. struct node_info *ni)
  99. {
  100. raw_ne->ino = cpu_to_le32(ni->ino);
  101. raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
  102. raw_ne->version = ni->version;
  103. }
  104. static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
  105. {
  106. return NM_I(sbi)->dirty_nat_cnt >= NM_I(sbi)->max_nid *
  107. NM_I(sbi)->dirty_nats_ratio / 100;
  108. }
  109. static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
  110. {
  111. return NM_I(sbi)->nat_cnt >= DEF_NAT_CACHE_THRESHOLD;
  112. }
  113. static inline bool excess_dirty_nodes(struct f2fs_sb_info *sbi)
  114. {
  115. return get_pages(sbi, F2FS_DIRTY_NODES) >= sbi->blocks_per_seg * 8;
  116. }
  117. enum mem_type {
  118. FREE_NIDS, /* indicates the free nid list */
  119. NAT_ENTRIES, /* indicates the cached nat entry */
  120. DIRTY_DENTS, /* indicates dirty dentry pages */
  121. INO_ENTRIES, /* indicates inode entries */
  122. EXTENT_CACHE, /* indicates extent cache */
  123. INMEM_PAGES, /* indicates inmemory pages */
  124. BASE_CHECK, /* check kernel status */
  125. };
  126. struct nat_entry_set {
  127. struct list_head set_list; /* link with other nat sets */
  128. struct list_head entry_list; /* link with dirty nat entries */
  129. nid_t set; /* set number*/
  130. unsigned int entry_cnt; /* the # of nat entries in set */
  131. };
  132. struct free_nid {
  133. struct list_head list; /* for free node id list */
  134. nid_t nid; /* node id */
  135. int state; /* in use or not: FREE_NID or PREALLOC_NID */
  136. };
  137. static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
  138. {
  139. struct f2fs_nm_info *nm_i = NM_I(sbi);
  140. struct free_nid *fnid;
  141. spin_lock(&nm_i->nid_list_lock);
  142. if (nm_i->nid_cnt[FREE_NID] <= 0) {
  143. spin_unlock(&nm_i->nid_list_lock);
  144. return;
  145. }
  146. fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
  147. *nid = fnid->nid;
  148. spin_unlock(&nm_i->nid_list_lock);
  149. }
  150. /*
  151. * inline functions
  152. */
  153. static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
  154. {
  155. struct f2fs_nm_info *nm_i = NM_I(sbi);
  156. #ifdef CONFIG_F2FS_CHECK_FS
  157. if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
  158. nm_i->bitmap_size))
  159. f2fs_bug_on(sbi, 1);
  160. #endif
  161. memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
  162. }
  163. static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
  164. {
  165. struct f2fs_nm_info *nm_i = NM_I(sbi);
  166. pgoff_t block_off;
  167. pgoff_t block_addr;
  168. /*
  169. * block_off = segment_off * 512 + off_in_segment
  170. * OLD = (segment_off * 512) * 2 + off_in_segment
  171. * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
  172. */
  173. block_off = NAT_BLOCK_OFFSET(start);
  174. block_addr = (pgoff_t)(nm_i->nat_blkaddr +
  175. (block_off << 1) -
  176. (block_off & (sbi->blocks_per_seg - 1)));
  177. if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
  178. block_addr += sbi->blocks_per_seg;
  179. return block_addr;
  180. }
  181. static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
  182. pgoff_t block_addr)
  183. {
  184. struct f2fs_nm_info *nm_i = NM_I(sbi);
  185. block_addr -= nm_i->nat_blkaddr;
  186. block_addr ^= 1 << sbi->log_blocks_per_seg;
  187. return block_addr + nm_i->nat_blkaddr;
  188. }
  189. static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
  190. {
  191. unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
  192. f2fs_change_bit(block_off, nm_i->nat_bitmap);
  193. #ifdef CONFIG_F2FS_CHECK_FS
  194. f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
  195. #endif
  196. }
  197. static inline nid_t ino_of_node(struct page *node_page)
  198. {
  199. struct f2fs_node *rn = F2FS_NODE(node_page);
  200. return le32_to_cpu(rn->footer.ino);
  201. }
  202. static inline nid_t nid_of_node(struct page *node_page)
  203. {
  204. struct f2fs_node *rn = F2FS_NODE(node_page);
  205. return le32_to_cpu(rn->footer.nid);
  206. }
  207. static inline unsigned int ofs_of_node(struct page *node_page)
  208. {
  209. struct f2fs_node *rn = F2FS_NODE(node_page);
  210. unsigned flag = le32_to_cpu(rn->footer.flag);
  211. return flag >> OFFSET_BIT_SHIFT;
  212. }
  213. static inline __u64 cpver_of_node(struct page *node_page)
  214. {
  215. struct f2fs_node *rn = F2FS_NODE(node_page);
  216. return le64_to_cpu(rn->footer.cp_ver);
  217. }
  218. static inline block_t next_blkaddr_of_node(struct page *node_page)
  219. {
  220. struct f2fs_node *rn = F2FS_NODE(node_page);
  221. return le32_to_cpu(rn->footer.next_blkaddr);
  222. }
  223. static inline void fill_node_footer(struct page *page, nid_t nid,
  224. nid_t ino, unsigned int ofs, bool reset)
  225. {
  226. struct f2fs_node *rn = F2FS_NODE(page);
  227. unsigned int old_flag = 0;
  228. if (reset)
  229. memset(rn, 0, sizeof(*rn));
  230. else
  231. old_flag = le32_to_cpu(rn->footer.flag);
  232. rn->footer.nid = cpu_to_le32(nid);
  233. rn->footer.ino = cpu_to_le32(ino);
  234. /* should remain old flag bits such as COLD_BIT_SHIFT */
  235. rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
  236. (old_flag & OFFSET_BIT_MASK));
  237. }
  238. static inline void copy_node_footer(struct page *dst, struct page *src)
  239. {
  240. struct f2fs_node *src_rn = F2FS_NODE(src);
  241. struct f2fs_node *dst_rn = F2FS_NODE(dst);
  242. memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
  243. }
  244. static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
  245. {
  246. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  247. struct f2fs_node *rn = F2FS_NODE(page);
  248. __u64 cp_ver = cur_cp_version(ckpt);
  249. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  250. cp_ver |= (cur_cp_crc(ckpt) << 32);
  251. rn->footer.cp_ver = cpu_to_le64(cp_ver);
  252. rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
  253. }
  254. static inline bool is_recoverable_dnode(struct page *page)
  255. {
  256. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  257. __u64 cp_ver = cur_cp_version(ckpt);
  258. /* Don't care crc part, if fsck.f2fs sets it. */
  259. if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
  260. return (cp_ver << 32) == (cpver_of_node(page) << 32);
  261. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  262. cp_ver |= (cur_cp_crc(ckpt) << 32);
  263. return cp_ver == cpver_of_node(page);
  264. }
  265. /*
  266. * f2fs assigns the following node offsets described as (num).
  267. * N = NIDS_PER_BLOCK
  268. *
  269. * Inode block (0)
  270. * |- direct node (1)
  271. * |- direct node (2)
  272. * |- indirect node (3)
  273. * | `- direct node (4 => 4 + N - 1)
  274. * |- indirect node (4 + N)
  275. * | `- direct node (5 + N => 5 + 2N - 1)
  276. * `- double indirect node (5 + 2N)
  277. * `- indirect node (6 + 2N)
  278. * `- direct node
  279. * ......
  280. * `- indirect node ((6 + 2N) + x(N + 1))
  281. * `- direct node
  282. * ......
  283. * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
  284. * `- direct node
  285. */
  286. static inline bool IS_DNODE(struct page *node_page)
  287. {
  288. unsigned int ofs = ofs_of_node(node_page);
  289. if (f2fs_has_xattr_block(ofs))
  290. return true;
  291. if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
  292. ofs == 5 + 2 * NIDS_PER_BLOCK)
  293. return false;
  294. if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
  295. ofs -= 6 + 2 * NIDS_PER_BLOCK;
  296. if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
  297. return false;
  298. }
  299. return true;
  300. }
  301. static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
  302. {
  303. struct f2fs_node *rn = F2FS_NODE(p);
  304. f2fs_wait_on_page_writeback(p, NODE, true, true);
  305. if (i)
  306. rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
  307. else
  308. rn->in.nid[off] = cpu_to_le32(nid);
  309. return set_page_dirty(p);
  310. }
  311. static inline nid_t get_nid(struct page *p, int off, bool i)
  312. {
  313. struct f2fs_node *rn = F2FS_NODE(p);
  314. if (i)
  315. return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
  316. return le32_to_cpu(rn->in.nid[off]);
  317. }
  318. /*
  319. * Coldness identification:
  320. * - Mark cold files in f2fs_inode_info
  321. * - Mark cold node blocks in their node footer
  322. * - Mark cold data pages in page cache
  323. */
  324. static inline int is_cold_data(struct page *page)
  325. {
  326. return PageChecked(page);
  327. }
  328. static inline void set_cold_data(struct page *page)
  329. {
  330. SetPageChecked(page);
  331. }
  332. static inline void clear_cold_data(struct page *page)
  333. {
  334. ClearPageChecked(page);
  335. }
  336. static inline int is_node(struct page *page, int type)
  337. {
  338. struct f2fs_node *rn = F2FS_NODE(page);
  339. return le32_to_cpu(rn->footer.flag) & (1 << type);
  340. }
  341. #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
  342. #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
  343. #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
  344. static inline int is_inline_node(struct page *page)
  345. {
  346. return PageChecked(page);
  347. }
  348. static inline void set_inline_node(struct page *page)
  349. {
  350. SetPageChecked(page);
  351. }
  352. static inline void clear_inline_node(struct page *page)
  353. {
  354. ClearPageChecked(page);
  355. }
  356. static inline void set_cold_node(struct page *page, bool is_dir)
  357. {
  358. struct f2fs_node *rn = F2FS_NODE(page);
  359. unsigned int flag = le32_to_cpu(rn->footer.flag);
  360. if (is_dir)
  361. flag &= ~(0x1 << COLD_BIT_SHIFT);
  362. else
  363. flag |= (0x1 << COLD_BIT_SHIFT);
  364. rn->footer.flag = cpu_to_le32(flag);
  365. }
  366. static inline void set_mark(struct page *page, int mark, int type)
  367. {
  368. struct f2fs_node *rn = F2FS_NODE(page);
  369. unsigned int flag = le32_to_cpu(rn->footer.flag);
  370. if (mark)
  371. flag |= (0x1 << type);
  372. else
  373. flag &= ~(0x1 << type);
  374. rn->footer.flag = cpu_to_le32(flag);
  375. #ifdef CONFIG_F2FS_CHECK_FS
  376. f2fs_inode_chksum_set(F2FS_P_SB(page), page);
  377. #endif
  378. }
  379. #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
  380. #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)