node.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434
  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. /* vector size for gang look-up from nat cache that consists of radix tree */
  27. #define NATVEC_SIZE 64
  28. #define SETVEC_SIZE 32
  29. /* return value for read_node_page */
  30. #define LOCKED_PAGE 1
  31. /* check pinned file's alignment status of physical blocks */
  32. #define FILE_NOT_ALIGNED 1
  33. /* For flag in struct node_info */
  34. enum {
  35. IS_CHECKPOINTED, /* is it checkpointed before? */
  36. HAS_FSYNCED_INODE, /* is the inode fsynced before? */
  37. HAS_LAST_FSYNC, /* has the latest node fsync mark? */
  38. IS_DIRTY, /* this nat entry is dirty? */
  39. IS_PREALLOC, /* nat entry is preallocated */
  40. };
  41. /*
  42. * For node information
  43. */
  44. struct node_info {
  45. nid_t nid; /* node id */
  46. nid_t ino; /* inode number of the node's owner */
  47. block_t blk_addr; /* block address of the node */
  48. unsigned char version; /* version of the node */
  49. unsigned char flag; /* for node information bits */
  50. };
  51. struct nat_entry {
  52. struct list_head list; /* for clean or dirty nat list */
  53. struct node_info ni; /* in-memory node information */
  54. };
  55. #define nat_get_nid(nat) ((nat)->ni.nid)
  56. #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
  57. #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
  58. #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
  59. #define nat_get_ino(nat) ((nat)->ni.ino)
  60. #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
  61. #define nat_get_version(nat) ((nat)->ni.version)
  62. #define nat_set_version(nat, v) ((nat)->ni.version = (v))
  63. #define inc_node_version(version) (++(version))
  64. static inline void copy_node_info(struct node_info *dst,
  65. struct node_info *src)
  66. {
  67. dst->nid = src->nid;
  68. dst->ino = src->ino;
  69. dst->blk_addr = src->blk_addr;
  70. dst->version = src->version;
  71. /* should not copy flag here */
  72. }
  73. static inline void set_nat_flag(struct nat_entry *ne,
  74. unsigned int type, bool set)
  75. {
  76. unsigned char mask = 0x01 << type;
  77. if (set)
  78. ne->ni.flag |= mask;
  79. else
  80. ne->ni.flag &= ~mask;
  81. }
  82. static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
  83. {
  84. unsigned char mask = 0x01 << type;
  85. return ne->ni.flag & mask;
  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. static inline bool excess_dirty_nodes(struct f2fs_sb_info *sbi)
  118. {
  119. return get_pages(sbi, F2FS_DIRTY_NODES) >= sbi->blocks_per_seg * 8;
  120. }
  121. enum mem_type {
  122. FREE_NIDS, /* indicates the free nid list */
  123. NAT_ENTRIES, /* indicates the cached nat entry */
  124. DIRTY_DENTS, /* indicates dirty dentry pages */
  125. INO_ENTRIES, /* indicates inode entries */
  126. EXTENT_CACHE, /* indicates extent cache */
  127. INMEM_PAGES, /* indicates inmemory pages */
  128. DISCARD_CACHE, /* indicates memory of cached discard cmds */
  129. COMPRESS_PAGE, /* indicates memory of cached compressed pages */
  130. BASE_CHECK, /* check kernel status */
  131. };
  132. struct nat_entry_set {
  133. struct list_head set_list; /* link with other nat sets */
  134. struct list_head entry_list; /* link with dirty nat entries */
  135. nid_t set; /* set number*/
  136. unsigned int entry_cnt; /* the # of nat entries in set */
  137. };
  138. struct free_nid {
  139. struct list_head list; /* for free node id list */
  140. nid_t nid; /* node id */
  141. int state; /* in use or not: FREE_NID or PREALLOC_NID */
  142. };
  143. static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
  144. {
  145. struct f2fs_nm_info *nm_i = NM_I(sbi);
  146. struct free_nid *fnid;
  147. spin_lock(&nm_i->nid_list_lock);
  148. if (nm_i->nid_cnt[FREE_NID] <= 0) {
  149. spin_unlock(&nm_i->nid_list_lock);
  150. return;
  151. }
  152. fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
  153. *nid = fnid->nid;
  154. spin_unlock(&nm_i->nid_list_lock);
  155. }
  156. /*
  157. * inline functions
  158. */
  159. static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
  160. {
  161. struct f2fs_nm_info *nm_i = NM_I(sbi);
  162. #ifdef CONFIG_F2FS_CHECK_FS
  163. if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
  164. nm_i->bitmap_size))
  165. f2fs_bug_on(sbi, 1);
  166. #endif
  167. memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
  168. }
  169. static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
  170. {
  171. struct f2fs_nm_info *nm_i = NM_I(sbi);
  172. pgoff_t block_off;
  173. pgoff_t block_addr;
  174. /*
  175. * block_off = segment_off * 512 + off_in_segment
  176. * OLD = (segment_off * 512) * 2 + off_in_segment
  177. * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
  178. */
  179. block_off = NAT_BLOCK_OFFSET(start);
  180. block_addr = (pgoff_t)(nm_i->nat_blkaddr +
  181. (block_off << 1) -
  182. (block_off & (sbi->blocks_per_seg - 1)));
  183. if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
  184. block_addr += sbi->blocks_per_seg;
  185. return block_addr;
  186. }
  187. static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
  188. pgoff_t block_addr)
  189. {
  190. struct f2fs_nm_info *nm_i = NM_I(sbi);
  191. block_addr -= nm_i->nat_blkaddr;
  192. block_addr ^= 1 << sbi->log_blocks_per_seg;
  193. return block_addr + nm_i->nat_blkaddr;
  194. }
  195. static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
  196. {
  197. unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
  198. f2fs_change_bit(block_off, nm_i->nat_bitmap);
  199. #ifdef CONFIG_F2FS_CHECK_FS
  200. f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
  201. #endif
  202. }
  203. static inline nid_t ino_of_node(struct page *node_page)
  204. {
  205. struct f2fs_node *rn = F2FS_NODE(node_page);
  206. return le32_to_cpu(rn->footer.ino);
  207. }
  208. static inline nid_t nid_of_node(struct page *node_page)
  209. {
  210. struct f2fs_node *rn = F2FS_NODE(node_page);
  211. return le32_to_cpu(rn->footer.nid);
  212. }
  213. static inline unsigned int ofs_of_node(struct page *node_page)
  214. {
  215. struct f2fs_node *rn = F2FS_NODE(node_page);
  216. unsigned flag = le32_to_cpu(rn->footer.flag);
  217. return flag >> OFFSET_BIT_SHIFT;
  218. }
  219. static inline __u64 cpver_of_node(struct page *node_page)
  220. {
  221. struct f2fs_node *rn = F2FS_NODE(node_page);
  222. return le64_to_cpu(rn->footer.cp_ver);
  223. }
  224. static inline block_t next_blkaddr_of_node(struct page *node_page)
  225. {
  226. struct f2fs_node *rn = F2FS_NODE(node_page);
  227. return le32_to_cpu(rn->footer.next_blkaddr);
  228. }
  229. static inline void fill_node_footer(struct page *page, nid_t nid,
  230. nid_t ino, unsigned int ofs, bool reset)
  231. {
  232. struct f2fs_node *rn = F2FS_NODE(page);
  233. unsigned int old_flag = 0;
  234. if (reset)
  235. memset(rn, 0, sizeof(*rn));
  236. else
  237. old_flag = le32_to_cpu(rn->footer.flag);
  238. rn->footer.nid = cpu_to_le32(nid);
  239. rn->footer.ino = cpu_to_le32(ino);
  240. /* should remain old flag bits such as COLD_BIT_SHIFT */
  241. rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
  242. (old_flag & OFFSET_BIT_MASK));
  243. }
  244. static inline void copy_node_footer(struct page *dst, struct page *src)
  245. {
  246. struct f2fs_node *src_rn = F2FS_NODE(src);
  247. struct f2fs_node *dst_rn = F2FS_NODE(dst);
  248. memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
  249. }
  250. static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
  251. {
  252. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  253. struct f2fs_node *rn = F2FS_NODE(page);
  254. __u64 cp_ver = cur_cp_version(ckpt);
  255. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  256. cp_ver |= (cur_cp_crc(ckpt) << 32);
  257. rn->footer.cp_ver = cpu_to_le64(cp_ver);
  258. rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
  259. }
  260. static inline bool is_recoverable_dnode(struct page *page)
  261. {
  262. struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
  263. __u64 cp_ver = cur_cp_version(ckpt);
  264. /* Don't care crc part, if fsck.f2fs sets it. */
  265. if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
  266. return (cp_ver << 32) == (cpver_of_node(page) << 32);
  267. if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
  268. cp_ver |= (cur_cp_crc(ckpt) << 32);
  269. return cp_ver == cpver_of_node(page);
  270. }
  271. /*
  272. * f2fs assigns the following node offsets described as (num).
  273. * N = NIDS_PER_BLOCK
  274. *
  275. * Inode block (0)
  276. * |- direct node (1)
  277. * |- direct node (2)
  278. * |- indirect node (3)
  279. * | `- direct node (4 => 4 + N - 1)
  280. * |- indirect node (4 + N)
  281. * | `- direct node (5 + N => 5 + 2N - 1)
  282. * `- double indirect node (5 + 2N)
  283. * `- indirect node (6 + 2N)
  284. * `- direct node
  285. * ......
  286. * `- indirect node ((6 + 2N) + x(N + 1))
  287. * `- direct node
  288. * ......
  289. * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
  290. * `- direct node
  291. */
  292. static inline bool IS_DNODE(struct page *node_page)
  293. {
  294. unsigned int ofs = ofs_of_node(node_page);
  295. if (f2fs_has_xattr_block(ofs))
  296. return true;
  297. if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
  298. ofs == 5 + 2 * NIDS_PER_BLOCK)
  299. return false;
  300. if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
  301. ofs -= 6 + 2 * NIDS_PER_BLOCK;
  302. if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
  303. return false;
  304. }
  305. return true;
  306. }
  307. static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
  308. {
  309. struct f2fs_node *rn = F2FS_NODE(p);
  310. f2fs_wait_on_page_writeback(p, NODE, true, true);
  311. if (i)
  312. rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
  313. else
  314. rn->in.nid[off] = cpu_to_le32(nid);
  315. return set_page_dirty(p);
  316. }
  317. static inline nid_t get_nid(struct page *p, int off, bool i)
  318. {
  319. struct f2fs_node *rn = F2FS_NODE(p);
  320. if (i)
  321. return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
  322. return le32_to_cpu(rn->in.nid[off]);
  323. }
  324. /*
  325. * Coldness identification:
  326. * - Mark cold files in f2fs_inode_info
  327. * - Mark cold node blocks in their node footer
  328. * - Mark cold data pages in page cache
  329. */
  330. static inline int is_node(struct page *page, int type)
  331. {
  332. struct f2fs_node *rn = F2FS_NODE(page);
  333. return le32_to_cpu(rn->footer.flag) & (1 << type);
  334. }
  335. #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
  336. #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
  337. #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
  338. static inline void set_cold_node(struct page *page, bool is_dir)
  339. {
  340. struct f2fs_node *rn = F2FS_NODE(page);
  341. unsigned int flag = le32_to_cpu(rn->footer.flag);
  342. if (is_dir)
  343. flag &= ~(0x1 << COLD_BIT_SHIFT);
  344. else
  345. flag |= (0x1 << COLD_BIT_SHIFT);
  346. rn->footer.flag = cpu_to_le32(flag);
  347. }
  348. static inline void set_mark(struct page *page, int mark, int type)
  349. {
  350. struct f2fs_node *rn = F2FS_NODE(page);
  351. unsigned int flag = le32_to_cpu(rn->footer.flag);
  352. if (mark)
  353. flag |= (0x1 << type);
  354. else
  355. flag &= ~(0x1 << type);
  356. rn->footer.flag = cpu_to_le32(flag);
  357. #ifdef CONFIG_F2FS_CHECK_FS
  358. f2fs_inode_chksum_set(F2FS_P_SB(page), page);
  359. #endif
  360. }
  361. #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
  362. #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)