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 11 /* node block offset on the NAT area dedicated to the given start node id */ 12 #define NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK) 13 14 /* # of pages to perform synchronous readahead before building free nids */ 15 #define FREE_NID_PAGES 8 16 #define MAX_FREE_NIDS (NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES) 17 18 /* size of free nid batch when shrinking */ 19 #define SHRINK_NID_BATCH_SIZE 8 20 21 #define DEF_RA_NID_PAGES 0 /* # of nid pages to be readaheaded */ 22 23 /* maximum readahead size for node during getting data blocks */ 24 #define MAX_RA_NODE 128 25 26 /* control the memory footprint threshold (10MB per 1GB ram) */ 27 #define DEF_RAM_THRESHOLD 1 28 29 /* control dirty nats ratio threshold (default: 10% over max nid count) */ 30 #define DEF_DIRTY_NAT_RATIO_THRESHOLD 10 31 /* control total # of nats */ 32 #define DEF_NAT_CACHE_THRESHOLD 100000 33 34 /* control total # of node writes used for roll-fowrad recovery */ 35 #define DEF_RF_NODE_BLOCKS 0 36 37 /* vector size for gang look-up from nat cache that consists of radix tree */ 38 #define NATVEC_SIZE 64 39 #define SETVEC_SIZE 32 40 41 /* return value for read_node_page */ 42 #define LOCKED_PAGE 1 43 44 /* check pinned file's alignment status of physical blocks */ 45 #define FILE_NOT_ALIGNED 1 46 47 /* For flag in struct node_info */ 48 enum { 49 IS_CHECKPOINTED, /* is it checkpointed before? */ 50 HAS_FSYNCED_INODE, /* is the inode fsynced before? */ 51 HAS_LAST_FSYNC, /* has the latest node fsync mark? */ 52 IS_DIRTY, /* this nat entry is dirty? */ 53 IS_PREALLOC, /* nat entry is preallocated */ 54 }; 55 56 /* 57 * For node information 58 */ 59 struct node_info { 60 nid_t nid; /* node id */ 61 nid_t ino; /* inode number of the node's owner */ 62 block_t blk_addr; /* block address of the node */ 63 unsigned char version; /* version of the node */ 64 unsigned char flag; /* for node information bits */ 65 }; 66 67 struct nat_entry { 68 struct list_head list; /* for clean or dirty nat list */ 69 struct node_info ni; /* in-memory node information */ 70 }; 71 72 #define nat_get_nid(nat) ((nat)->ni.nid) 73 #define nat_set_nid(nat, n) ((nat)->ni.nid = (n)) 74 #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr) 75 #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b)) 76 #define nat_get_ino(nat) ((nat)->ni.ino) 77 #define nat_set_ino(nat, i) ((nat)->ni.ino = (i)) 78 #define nat_get_version(nat) ((nat)->ni.version) 79 #define nat_set_version(nat, v) ((nat)->ni.version = (v)) 80 81 #define inc_node_version(version) (++(version)) 82 83 static inline void copy_node_info(struct node_info *dst, 84 struct node_info *src) 85 { 86 dst->nid = src->nid; 87 dst->ino = src->ino; 88 dst->blk_addr = src->blk_addr; 89 dst->version = src->version; 90 /* should not copy flag here */ 91 } 92 93 static inline void set_nat_flag(struct nat_entry *ne, 94 unsigned int type, bool set) 95 { 96 unsigned char mask = 0x01 << type; 97 if (set) 98 ne->ni.flag |= mask; 99 else 100 ne->ni.flag &= ~mask; 101 } 102 103 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type) 104 { 105 unsigned char mask = 0x01 << type; 106 return ne->ni.flag & mask; 107 } 108 109 static inline void nat_reset_flag(struct nat_entry *ne) 110 { 111 /* these states can be set only after checkpoint was done */ 112 set_nat_flag(ne, IS_CHECKPOINTED, true); 113 set_nat_flag(ne, HAS_FSYNCED_INODE, false); 114 set_nat_flag(ne, HAS_LAST_FSYNC, true); 115 } 116 117 static inline void node_info_from_raw_nat(struct node_info *ni, 118 struct f2fs_nat_entry *raw_ne) 119 { 120 ni->ino = le32_to_cpu(raw_ne->ino); 121 ni->blk_addr = le32_to_cpu(raw_ne->block_addr); 122 ni->version = raw_ne->version; 123 } 124 125 static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne, 126 struct node_info *ni) 127 { 128 raw_ne->ino = cpu_to_le32(ni->ino); 129 raw_ne->block_addr = cpu_to_le32(ni->blk_addr); 130 raw_ne->version = ni->version; 131 } 132 133 static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi) 134 { 135 return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid * 136 NM_I(sbi)->dirty_nats_ratio / 100; 137 } 138 139 static inline bool excess_cached_nats(struct f2fs_sb_info *sbi) 140 { 141 return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD; 142 } 143 144 enum mem_type { 145 FREE_NIDS, /* indicates the free nid list */ 146 NAT_ENTRIES, /* indicates the cached nat entry */ 147 DIRTY_DENTS, /* indicates dirty dentry pages */ 148 INO_ENTRIES, /* indicates inode entries */ 149 EXTENT_CACHE, /* indicates extent cache */ 150 INMEM_PAGES, /* indicates inmemory pages */ 151 DISCARD_CACHE, /* indicates memory of cached discard cmds */ 152 COMPRESS_PAGE, /* indicates memory of cached compressed pages */ 153 BASE_CHECK, /* check kernel status */ 154 }; 155 156 struct nat_entry_set { 157 struct list_head set_list; /* link with other nat sets */ 158 struct list_head entry_list; /* link with dirty nat entries */ 159 nid_t set; /* set number*/ 160 unsigned int entry_cnt; /* the # of nat entries in set */ 161 }; 162 163 struct free_nid { 164 struct list_head list; /* for free node id list */ 165 nid_t nid; /* node id */ 166 int state; /* in use or not: FREE_NID or PREALLOC_NID */ 167 }; 168 169 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid) 170 { 171 struct f2fs_nm_info *nm_i = NM_I(sbi); 172 struct free_nid *fnid; 173 174 spin_lock(&nm_i->nid_list_lock); 175 if (nm_i->nid_cnt[FREE_NID] <= 0) { 176 spin_unlock(&nm_i->nid_list_lock); 177 return; 178 } 179 fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list); 180 *nid = fnid->nid; 181 spin_unlock(&nm_i->nid_list_lock); 182 } 183 184 /* 185 * inline functions 186 */ 187 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr) 188 { 189 struct f2fs_nm_info *nm_i = NM_I(sbi); 190 191 #ifdef CONFIG_F2FS_CHECK_FS 192 if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir, 193 nm_i->bitmap_size)) 194 f2fs_bug_on(sbi, 1); 195 #endif 196 memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size); 197 } 198 199 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start) 200 { 201 struct f2fs_nm_info *nm_i = NM_I(sbi); 202 pgoff_t block_off; 203 pgoff_t block_addr; 204 205 /* 206 * block_off = segment_off * 512 + off_in_segment 207 * OLD = (segment_off * 512) * 2 + off_in_segment 208 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment 209 */ 210 block_off = NAT_BLOCK_OFFSET(start); 211 212 block_addr = (pgoff_t)(nm_i->nat_blkaddr + 213 (block_off << 1) - 214 (block_off & (sbi->blocks_per_seg - 1))); 215 216 if (f2fs_test_bit(block_off, nm_i->nat_bitmap)) 217 block_addr += sbi->blocks_per_seg; 218 219 return block_addr; 220 } 221 222 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi, 223 pgoff_t block_addr) 224 { 225 struct f2fs_nm_info *nm_i = NM_I(sbi); 226 227 block_addr -= nm_i->nat_blkaddr; 228 block_addr ^= 1 << sbi->log_blocks_per_seg; 229 return block_addr + nm_i->nat_blkaddr; 230 } 231 232 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid) 233 { 234 unsigned int block_off = NAT_BLOCK_OFFSET(start_nid); 235 236 f2fs_change_bit(block_off, nm_i->nat_bitmap); 237 #ifdef CONFIG_F2FS_CHECK_FS 238 f2fs_change_bit(block_off, nm_i->nat_bitmap_mir); 239 #endif 240 } 241 242 static inline nid_t ino_of_node(struct page *node_page) 243 { 244 struct f2fs_node *rn = F2FS_NODE(node_page); 245 return le32_to_cpu(rn->footer.ino); 246 } 247 248 static inline nid_t nid_of_node(struct page *node_page) 249 { 250 struct f2fs_node *rn = F2FS_NODE(node_page); 251 return le32_to_cpu(rn->footer.nid); 252 } 253 254 static inline unsigned int ofs_of_node(struct page *node_page) 255 { 256 struct f2fs_node *rn = F2FS_NODE(node_page); 257 unsigned flag = le32_to_cpu(rn->footer.flag); 258 return flag >> OFFSET_BIT_SHIFT; 259 } 260 261 static inline __u64 cpver_of_node(struct page *node_page) 262 { 263 struct f2fs_node *rn = F2FS_NODE(node_page); 264 return le64_to_cpu(rn->footer.cp_ver); 265 } 266 267 static inline block_t next_blkaddr_of_node(struct page *node_page) 268 { 269 struct f2fs_node *rn = F2FS_NODE(node_page); 270 return le32_to_cpu(rn->footer.next_blkaddr); 271 } 272 273 static inline void fill_node_footer(struct page *page, nid_t nid, 274 nid_t ino, unsigned int ofs, bool reset) 275 { 276 struct f2fs_node *rn = F2FS_NODE(page); 277 unsigned int old_flag = 0; 278 279 if (reset) 280 memset(rn, 0, sizeof(*rn)); 281 else 282 old_flag = le32_to_cpu(rn->footer.flag); 283 284 rn->footer.nid = cpu_to_le32(nid); 285 rn->footer.ino = cpu_to_le32(ino); 286 287 /* should remain old flag bits such as COLD_BIT_SHIFT */ 288 rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) | 289 (old_flag & OFFSET_BIT_MASK)); 290 } 291 292 static inline void copy_node_footer(struct page *dst, struct page *src) 293 { 294 struct f2fs_node *src_rn = F2FS_NODE(src); 295 struct f2fs_node *dst_rn = F2FS_NODE(dst); 296 memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer)); 297 } 298 299 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr) 300 { 301 struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page)); 302 struct f2fs_node *rn = F2FS_NODE(page); 303 __u64 cp_ver = cur_cp_version(ckpt); 304 305 if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG)) 306 cp_ver |= (cur_cp_crc(ckpt) << 32); 307 308 rn->footer.cp_ver = cpu_to_le64(cp_ver); 309 rn->footer.next_blkaddr = cpu_to_le32(blkaddr); 310 } 311 312 static inline bool is_recoverable_dnode(struct page *page) 313 { 314 struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page)); 315 __u64 cp_ver = cur_cp_version(ckpt); 316 317 /* Don't care crc part, if fsck.f2fs sets it. */ 318 if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG)) 319 return (cp_ver << 32) == (cpver_of_node(page) << 32); 320 321 if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG)) 322 cp_ver |= (cur_cp_crc(ckpt) << 32); 323 324 return cp_ver == cpver_of_node(page); 325 } 326 327 /* 328 * f2fs assigns the following node offsets described as (num). 329 * N = NIDS_PER_BLOCK 330 * 331 * Inode block (0) 332 * |- direct node (1) 333 * |- direct node (2) 334 * |- indirect node (3) 335 * | `- direct node (4 => 4 + N - 1) 336 * |- indirect node (4 + N) 337 * | `- direct node (5 + N => 5 + 2N - 1) 338 * `- double indirect node (5 + 2N) 339 * `- indirect node (6 + 2N) 340 * `- direct node 341 * ...... 342 * `- indirect node ((6 + 2N) + x(N + 1)) 343 * `- direct node 344 * ...... 345 * `- indirect node ((6 + 2N) + (N - 1)(N + 1)) 346 * `- direct node 347 */ 348 static inline bool IS_DNODE(struct page *node_page) 349 { 350 unsigned int ofs = ofs_of_node(node_page); 351 352 if (f2fs_has_xattr_block(ofs)) 353 return true; 354 355 if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK || 356 ofs == 5 + 2 * NIDS_PER_BLOCK) 357 return false; 358 if (ofs >= 6 + 2 * NIDS_PER_BLOCK) { 359 ofs -= 6 + 2 * NIDS_PER_BLOCK; 360 if (!((long int)ofs % (NIDS_PER_BLOCK + 1))) 361 return false; 362 } 363 return true; 364 } 365 366 static inline int set_nid(struct page *p, int off, nid_t nid, bool i) 367 { 368 struct f2fs_node *rn = F2FS_NODE(p); 369 370 f2fs_wait_on_page_writeback(p, NODE, true, true); 371 372 if (i) 373 rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid); 374 else 375 rn->in.nid[off] = cpu_to_le32(nid); 376 return set_page_dirty(p); 377 } 378 379 static inline nid_t get_nid(struct page *p, int off, bool i) 380 { 381 struct f2fs_node *rn = F2FS_NODE(p); 382 383 if (i) 384 return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]); 385 return le32_to_cpu(rn->in.nid[off]); 386 } 387 388 /* 389 * Coldness identification: 390 * - Mark cold files in f2fs_inode_info 391 * - Mark cold node blocks in their node footer 392 * - Mark cold data pages in page cache 393 */ 394 395 static inline int is_node(struct page *page, int type) 396 { 397 struct f2fs_node *rn = F2FS_NODE(page); 398 return le32_to_cpu(rn->footer.flag) & (1 << type); 399 } 400 401 #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT) 402 #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT) 403 #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT) 404 405 static inline void set_cold_node(struct page *page, bool is_dir) 406 { 407 struct f2fs_node *rn = F2FS_NODE(page); 408 unsigned int flag = le32_to_cpu(rn->footer.flag); 409 410 if (is_dir) 411 flag &= ~(0x1 << COLD_BIT_SHIFT); 412 else 413 flag |= (0x1 << COLD_BIT_SHIFT); 414 rn->footer.flag = cpu_to_le32(flag); 415 } 416 417 static inline void set_mark(struct page *page, int mark, int type) 418 { 419 struct f2fs_node *rn = F2FS_NODE(page); 420 unsigned int flag = le32_to_cpu(rn->footer.flag); 421 if (mark) 422 flag |= (0x1 << type); 423 else 424 flag &= ~(0x1 << type); 425 rn->footer.flag = cpu_to_le32(flag); 426 427 #ifdef CONFIG_F2FS_CHECK_FS 428 f2fs_inode_chksum_set(F2FS_P_SB(page), page); 429 #endif 430 } 431 #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT) 432 #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT) 433