node.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430
  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. /* size of free nid batch when shrinking */
  16. #define SHRINK_NID_BATCH_SIZE 8
  17. #define DEF_RA_NID_PAGES 0 /* # of nid pages to be readaheaded */
  18. /* maximum readahead size for node during getting data blocks */
  19. #define MAX_RA_NODE 128
  20. /* control the memory footprint threshold (10MB per 1GB ram) */
  21. #define DEF_RAM_THRESHOLD 1
  22. /* control dirty nats ratio threshold (default: 10% over max nid count) */
  23. #define DEF_DIRTY_NAT_RATIO_THRESHOLD 10
  24. /* control total # of nats */
  25. #define DEF_NAT_CACHE_THRESHOLD 100000
  26. /* control total # of node writes used for roll-fowrad recovery */
  27. #define DEF_RF_NODE_BLOCKS 0
  28. /* vector size for gang look-up from nat cache that consists of radix tree */
  29. #define NATVEC_SIZE 64
  30. #define SETVEC_SIZE 32
  31. /* return value for read_node_page */
  32. #define LOCKED_PAGE 1
  33. /* check pinned file's alignment status of physical blocks */
  34. #define FILE_NOT_ALIGNED 1
  35. /* For flag in struct node_info */
  36. enum {
  37. IS_CHECKPOINTED, /* is it checkpointed before? */
  38. HAS_FSYNCED_INODE, /* is the inode fsynced before? */
  39. HAS_LAST_FSYNC, /* has the latest node fsync mark? */
  40. IS_DIRTY, /* this nat entry is dirty? */
  41. IS_PREALLOC, /* nat entry is preallocated */
  42. };
  43. /*
  44. * For node information
  45. */
  46. struct node_info {
  47. nid_t nid; /* node id */
  48. nid_t ino; /* inode number of the node's owner */
  49. block_t blk_addr; /* block address of the node */
  50. unsigned char version; /* version of the node */
  51. unsigned char flag; /* for node information bits */
  52. };
  53. struct nat_entry {
  54. struct list_head list; /* for clean or dirty nat list */
  55. struct node_info ni; /* in-memory node information */
  56. };
  57. #define nat_get_nid(nat) ((nat)->ni.nid)
  58. #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
  59. #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
  60. #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
  61. #define nat_get_ino(nat) ((nat)->ni.ino)
  62. #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
  63. #define nat_get_version(nat) ((nat)->ni.version)
  64. #define nat_set_version(nat, v) ((nat)->ni.version = (v))
  65. #define inc_node_version(version) (++(version))
  66. static inline void copy_node_info(struct node_info *dst,
  67. struct node_info *src)
  68. {
  69. dst->nid = src->nid;
  70. dst->ino = src->ino;
  71. dst->blk_addr = src->blk_addr;
  72. dst->version = src->version;
  73. /* should not copy flag here */
  74. }
  75. static inline void set_nat_flag(struct nat_entry *ne,
  76. unsigned int type, bool set)
  77. {
  78. if (set)
  79. ne->ni.flag |= BIT(type);
  80. else
  81. ne->ni.flag &= ~BIT(type);
  82. }
  83. static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
  84. {
  85. return ne->ni.flag & BIT(type);
  86. }
  87. static inline void nat_reset_flag(struct nat_entry *ne)
  88. {
  89. /* these states can be set only after checkpoint was done */
  90. set_nat_flag(ne, IS_CHECKPOINTED, true);
  91. set_nat_flag(ne, HAS_FSYNCED_INODE, false);
  92. set_nat_flag(ne, HAS_LAST_FSYNC, true);
  93. }
  94. static inline void node_info_from_raw_nat(struct node_info *ni,
  95. struct f2fs_nat_entry *raw_ne)
  96. {
  97. ni->ino = le32_to_cpu(raw_ne->ino);
  98. ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
  99. ni->version = raw_ne->version;
  100. }
  101. static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
  102. struct node_info *ni)
  103. {
  104. raw_ne->ino = cpu_to_le32(ni->ino);
  105. raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
  106. raw_ne->version = ni->version;
  107. }
  108. static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
  109. {
  110. return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid *
  111. NM_I(sbi)->dirty_nats_ratio / 100;
  112. }
  113. static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
  114. {
  115. return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD;
  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. READ_EXTENT_CACHE, /* indicates read extent cache */
  123. AGE_EXTENT_CACHE, /* indicates age extent cache */
  124. DISCARD_CACHE, /* indicates memory of cached discard cmds */
  125. COMPRESS_PAGE, /* indicates memory of cached compressed pages */
  126. BASE_CHECK, /* check kernel status */
  127. };
  128. struct nat_entry_set {
  129. struct list_head set_list; /* link with other nat sets */
  130. struct list_head entry_list; /* link with dirty nat entries */
  131. nid_t set; /* set number*/
  132. unsigned int entry_cnt; /* the # of nat entries in set */
  133. };
  134. struct free_nid {
  135. struct list_head list; /* for free node id list */
  136. nid_t nid; /* node id */
  137. int state; /* in use or not: FREE_NID or PREALLOC_NID */
  138. };
  139. static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
  140. {
  141. struct f2fs_nm_info *nm_i = NM_I(sbi);
  142. struct free_nid *fnid;
  143. spin_lock(&nm_i->nid_list_lock);
  144. if (nm_i->nid_cnt[FREE_NID] <= 0) {
  145. spin_unlock(&nm_i->nid_list_lock);
  146. return;
  147. }
  148. fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
  149. *nid = fnid->nid;
  150. spin_unlock(&nm_i->nid_list_lock);
  151. }
  152. /*
  153. * inline functions
  154. */
  155. static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
  156. {
  157. struct f2fs_nm_info *nm_i = NM_I(sbi);
  158. #ifdef CONFIG_F2FS_CHECK_FS
  159. if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
  160. nm_i->bitmap_size))
  161. f2fs_bug_on(sbi, 1);
  162. #endif
  163. memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
  164. }
  165. static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
  166. {
  167. struct f2fs_nm_info *nm_i = NM_I(sbi);
  168. pgoff_t block_off;
  169. pgoff_t block_addr;
  170. /*
  171. * block_off = segment_off * 512 + off_in_segment
  172. * OLD = (segment_off * 512) * 2 + off_in_segment
  173. * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
  174. */
  175. block_off = NAT_BLOCK_OFFSET(start);
  176. block_addr = (pgoff_t)(nm_i->nat_blkaddr +
  177. (block_off << 1) -
  178. (block_off & (sbi->blocks_per_seg - 1)));
  179. if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
  180. block_addr += sbi->blocks_per_seg;
  181. return block_addr;
  182. }
  183. static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
  184. pgoff_t block_addr)
  185. {
  186. struct f2fs_nm_info *nm_i = NM_I(sbi);
  187. block_addr -= nm_i->nat_blkaddr;
  188. block_addr ^= BIT(sbi->log_blocks_per_seg);
  189. return block_addr + nm_i->nat_blkaddr;
  190. }
  191. static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
  192. {
  193. unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
  194. f2fs_change_bit(block_off, nm_i->nat_bitmap);
  195. #ifdef CONFIG_F2FS_CHECK_FS
  196. f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
  197. #endif
  198. }
  199. static inline nid_t ino_of_node(struct page *node_page)
  200. {
  201. struct f2fs_node *rn = F2FS_NODE(node_page);
  202. return le32_to_cpu(rn->footer.ino);
  203. }
  204. static inline nid_t nid_of_node(struct page *node_page)
  205. {
  206. struct f2fs_node *rn = F2FS_NODE(node_page);
  207. return le32_to_cpu(rn->footer.nid);
  208. }
  209. static inline unsigned int ofs_of_node(struct page *node_page)
  210. {
  211. struct f2fs_node *rn = F2FS_NODE(node_page);
  212. unsigned flag = le32_to_cpu(rn->footer.flag);
  213. return flag >> OFFSET_BIT_SHIFT;
  214. }
  215. static inline __u64 cpver_of_node(struct page *node_page)
  216. {
  217. struct f2fs_node *rn = F2FS_NODE(node_page);
  218. return le64_to_cpu(rn->footer.cp_ver);
  219. }
  220. static inline block_t next_blkaddr_of_node(struct page *node_page)
  221. {
  222. struct f2fs_node *rn = F2FS_NODE(node_page);
  223. return le32_to_cpu(rn->footer.next_blkaddr);
  224. }
  225. static inline void fill_node_footer(struct page *page, nid_t nid,
  226. nid_t ino, unsigned int ofs, bool reset)
  227. {
  228. struct f2fs_node *rn = F2FS_NODE(page);
  229. unsigned int old_flag = 0;
  230. if (reset)
  231. memset(rn, 0, sizeof(*rn));
  232. else
  233. old_flag = le32_to_cpu(rn->footer.flag);
  234. rn->footer.nid = cpu_to_le32(nid);
  235. rn->footer.ino = cpu_to_le32(ino);
  236. /* should remain old flag bits such as COLD_BIT_SHIFT */
  237. rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
  238. (old_flag & OFFSET_BIT_MASK));
  239. }
  240. static inline void copy_node_footer(struct page *dst, struct page *src)
  241. {
  242. struct f2fs_node *src_rn = F2FS_NODE(src);
  243. struct f2fs_node *dst_rn = F2FS_NODE(dst);
  244. memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
  245. }
  246. static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
  247. {
  248. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  249. struct f2fs_node *rn = F2FS_NODE(page);
  250. __u64 cp_ver = cur_cp_version(ckpt);
  251. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  252. cp_ver |= (cur_cp_crc(ckpt) << 32);
  253. rn->footer.cp_ver = cpu_to_le64(cp_ver);
  254. rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
  255. }
  256. static inline bool is_recoverable_dnode(struct page *page)
  257. {
  258. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  259. __u64 cp_ver = cur_cp_version(ckpt);
  260. /* Don't care crc part, if fsck.f2fs sets it. */
  261. if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
  262. return (cp_ver << 32) == (cpver_of_node(page) << 32);
  263. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  264. cp_ver |= (cur_cp_crc(ckpt) << 32);
  265. return cp_ver == cpver_of_node(page);
  266. }
  267. /*
  268. * f2fs assigns the following node offsets described as (num).
  269. * N = NIDS_PER_BLOCK
  270. *
  271. * Inode block (0)
  272. * |- direct node (1)
  273. * |- direct node (2)
  274. * |- indirect node (3)
  275. * | `- direct node (4 => 4 + N - 1)
  276. * |- indirect node (4 + N)
  277. * | `- direct node (5 + N => 5 + 2N - 1)
  278. * `- double indirect node (5 + 2N)
  279. * `- indirect node (6 + 2N)
  280. * `- direct node
  281. * ......
  282. * `- indirect node ((6 + 2N) + x(N + 1))
  283. * `- direct node
  284. * ......
  285. * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
  286. * `- direct node
  287. */
  288. static inline bool IS_DNODE(struct page *node_page)
  289. {
  290. unsigned int ofs = ofs_of_node(node_page);
  291. if (f2fs_has_xattr_block(ofs))
  292. return true;
  293. if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
  294. ofs == 5 + 2 * NIDS_PER_BLOCK)
  295. return false;
  296. if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
  297. ofs -= 6 + 2 * NIDS_PER_BLOCK;
  298. if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
  299. return false;
  300. }
  301. return true;
  302. }
  303. static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
  304. {
  305. struct f2fs_node *rn = F2FS_NODE(p);
  306. f2fs_wait_on_page_writeback(p, NODE, true, true);
  307. if (i)
  308. rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
  309. else
  310. rn->in.nid[off] = cpu_to_le32(nid);
  311. return set_page_dirty(p);
  312. }
  313. static inline nid_t get_nid(struct page *p, int off, bool i)
  314. {
  315. struct f2fs_node *rn = F2FS_NODE(p);
  316. if (i)
  317. return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
  318. return le32_to_cpu(rn->in.nid[off]);
  319. }
  320. /*
  321. * Coldness identification:
  322. * - Mark cold files in f2fs_inode_info
  323. * - Mark cold node blocks in their node footer
  324. * - Mark cold data pages in page cache
  325. */
  326. static inline int is_node(struct page *page, int type)
  327. {
  328. struct f2fs_node *rn = F2FS_NODE(page);
  329. return le32_to_cpu(rn->footer.flag) & BIT(type);
  330. }
  331. #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
  332. #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
  333. #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
  334. static inline void set_cold_node(struct page *page, bool is_dir)
  335. {
  336. struct f2fs_node *rn = F2FS_NODE(page);
  337. unsigned int flag = le32_to_cpu(rn->footer.flag);
  338. if (is_dir)
  339. flag &= ~BIT(COLD_BIT_SHIFT);
  340. else
  341. flag |= BIT(COLD_BIT_SHIFT);
  342. rn->footer.flag = cpu_to_le32(flag);
  343. }
  344. static inline void set_mark(struct page *page, int mark, int type)
  345. {
  346. struct f2fs_node *rn = F2FS_NODE(page);
  347. unsigned int flag = le32_to_cpu(rn->footer.flag);
  348. if (mark)
  349. flag |= BIT(type);
  350. else
  351. flag &= ~BIT(type);
  352. rn->footer.flag = cpu_to_le32(flag);
  353. #ifdef CONFIG_F2FS_CHECK_FS
  354. f2fs_inode_chksum_set(F2FS_P_SB(page), page);
  355. #endif
  356. }
  357. #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
  358. #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)