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 if (set) 97 ne->ni.flag |= BIT(type); 98 else 99 ne->ni.flag &= ~BIT(type); 100 } 101 102 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type) 103 { 104 return ne->ni.flag & BIT(type); 105 } 106 107 static inline void nat_reset_flag(struct nat_entry *ne) 108 { 109 /* these states can be set only after checkpoint was done */ 110 set_nat_flag(ne, IS_CHECKPOINTED, true); 111 set_nat_flag(ne, HAS_FSYNCED_INODE, false); 112 set_nat_flag(ne, HAS_LAST_FSYNC, true); 113 } 114 115 static inline void node_info_from_raw_nat(struct node_info *ni, 116 struct f2fs_nat_entry *raw_ne) 117 { 118 ni->ino = le32_to_cpu(raw_ne->ino); 119 ni->blk_addr = le32_to_cpu(raw_ne->block_addr); 120 ni->version = raw_ne->version; 121 } 122 123 static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne, 124 struct node_info *ni) 125 { 126 raw_ne->ino = cpu_to_le32(ni->ino); 127 raw_ne->block_addr = cpu_to_le32(ni->blk_addr); 128 raw_ne->version = ni->version; 129 } 130 131 static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi) 132 { 133 return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid * 134 NM_I(sbi)->dirty_nats_ratio / 100; 135 } 136 137 static inline bool excess_cached_nats(struct f2fs_sb_info *sbi) 138 { 139 return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD; 140 } 141 142 enum mem_type { 143 FREE_NIDS, /* indicates the free nid list */ 144 NAT_ENTRIES, /* indicates the cached nat entry */ 145 DIRTY_DENTS, /* indicates dirty dentry pages */ 146 INO_ENTRIES, /* indicates inode entries */ 147 READ_EXTENT_CACHE, /* indicates read extent cache */ 148 AGE_EXTENT_CACHE, /* indicates age extent cache */ 149 DISCARD_CACHE, /* indicates memory of cached discard cmds */ 150 COMPRESS_PAGE, /* indicates memory of cached compressed pages */ 151 BASE_CHECK, /* check kernel status */ 152 }; 153 154 struct nat_entry_set { 155 struct list_head set_list; /* link with other nat sets */ 156 struct list_head entry_list; /* link with dirty nat entries */ 157 nid_t set; /* set number*/ 158 unsigned int entry_cnt; /* the # of nat entries in set */ 159 }; 160 161 struct free_nid { 162 struct list_head list; /* for free node id list */ 163 nid_t nid; /* node id */ 164 int state; /* in use or not: FREE_NID or PREALLOC_NID */ 165 }; 166 167 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid) 168 { 169 struct f2fs_nm_info *nm_i = NM_I(sbi); 170 struct free_nid *fnid; 171 172 spin_lock(&nm_i->nid_list_lock); 173 if (nm_i->nid_cnt[FREE_NID] <= 0) { 174 spin_unlock(&nm_i->nid_list_lock); 175 return; 176 } 177 fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list); 178 *nid = fnid->nid; 179 spin_unlock(&nm_i->nid_list_lock); 180 } 181 182 /* 183 * inline functions 184 */ 185 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr) 186 { 187 struct f2fs_nm_info *nm_i = NM_I(sbi); 188 189 #ifdef CONFIG_F2FS_CHECK_FS 190 if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir, 191 nm_i->bitmap_size)) 192 f2fs_bug_on(sbi, 1); 193 #endif 194 memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size); 195 } 196 197 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start) 198 { 199 struct f2fs_nm_info *nm_i = NM_I(sbi); 200 pgoff_t block_off; 201 pgoff_t block_addr; 202 203 /* 204 * block_off = segment_off * 512 + off_in_segment 205 * OLD = (segment_off * 512) * 2 + off_in_segment 206 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment 207 */ 208 block_off = NAT_BLOCK_OFFSET(start); 209 210 block_addr = (pgoff_t)(nm_i->nat_blkaddr + 211 (block_off << 1) - 212 (block_off & (sbi->blocks_per_seg - 1))); 213 214 if (f2fs_test_bit(block_off, nm_i->nat_bitmap)) 215 block_addr += sbi->blocks_per_seg; 216 217 return block_addr; 218 } 219 220 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi, 221 pgoff_t block_addr) 222 { 223 struct f2fs_nm_info *nm_i = NM_I(sbi); 224 225 block_addr -= nm_i->nat_blkaddr; 226 block_addr ^= BIT(sbi->log_blocks_per_seg); 227 return block_addr + nm_i->nat_blkaddr; 228 } 229 230 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid) 231 { 232 unsigned int block_off = NAT_BLOCK_OFFSET(start_nid); 233 234 f2fs_change_bit(block_off, nm_i->nat_bitmap); 235 #ifdef CONFIG_F2FS_CHECK_FS 236 f2fs_change_bit(block_off, nm_i->nat_bitmap_mir); 237 #endif 238 } 239 240 static inline nid_t ino_of_node(struct page *node_page) 241 { 242 struct f2fs_node *rn = F2FS_NODE(node_page); 243 return le32_to_cpu(rn->footer.ino); 244 } 245 246 static inline nid_t nid_of_node(struct page *node_page) 247 { 248 struct f2fs_node *rn = F2FS_NODE(node_page); 249 return le32_to_cpu(rn->footer.nid); 250 } 251 252 static inline unsigned int ofs_of_node(struct page *node_page) 253 { 254 struct f2fs_node *rn = F2FS_NODE(node_page); 255 unsigned flag = le32_to_cpu(rn->footer.flag); 256 return flag >> OFFSET_BIT_SHIFT; 257 } 258 259 static inline __u64 cpver_of_node(struct page *node_page) 260 { 261 struct f2fs_node *rn = F2FS_NODE(node_page); 262 return le64_to_cpu(rn->footer.cp_ver); 263 } 264 265 static inline block_t next_blkaddr_of_node(struct page *node_page) 266 { 267 struct f2fs_node *rn = F2FS_NODE(node_page); 268 return le32_to_cpu(rn->footer.next_blkaddr); 269 } 270 271 static inline void fill_node_footer(struct page *page, nid_t nid, 272 nid_t ino, unsigned int ofs, bool reset) 273 { 274 struct f2fs_node *rn = F2FS_NODE(page); 275 unsigned int old_flag = 0; 276 277 if (reset) 278 memset(rn, 0, sizeof(*rn)); 279 else 280 old_flag = le32_to_cpu(rn->footer.flag); 281 282 rn->footer.nid = cpu_to_le32(nid); 283 rn->footer.ino = cpu_to_le32(ino); 284 285 /* should remain old flag bits such as COLD_BIT_SHIFT */ 286 rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) | 287 (old_flag & OFFSET_BIT_MASK)); 288 } 289 290 static inline void copy_node_footer(struct page *dst, struct page *src) 291 { 292 struct f2fs_node *src_rn = F2FS_NODE(src); 293 struct f2fs_node *dst_rn = F2FS_NODE(dst); 294 memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer)); 295 } 296 297 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr) 298 { 299 struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page)); 300 struct f2fs_node *rn = F2FS_NODE(page); 301 __u64 cp_ver = cur_cp_version(ckpt); 302 303 if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG)) 304 cp_ver |= (cur_cp_crc(ckpt) << 32); 305 306 rn->footer.cp_ver = cpu_to_le64(cp_ver); 307 rn->footer.next_blkaddr = cpu_to_le32(blkaddr); 308 } 309 310 static inline bool is_recoverable_dnode(struct page *page) 311 { 312 struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page)); 313 __u64 cp_ver = cur_cp_version(ckpt); 314 315 /* Don't care crc part, if fsck.f2fs sets it. */ 316 if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG)) 317 return (cp_ver << 32) == (cpver_of_node(page) << 32); 318 319 if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG)) 320 cp_ver |= (cur_cp_crc(ckpt) << 32); 321 322 return cp_ver == cpver_of_node(page); 323 } 324 325 /* 326 * f2fs assigns the following node offsets described as (num). 327 * N = NIDS_PER_BLOCK 328 * 329 * Inode block (0) 330 * |- direct node (1) 331 * |- direct node (2) 332 * |- indirect node (3) 333 * | `- direct node (4 => 4 + N - 1) 334 * |- indirect node (4 + N) 335 * | `- direct node (5 + N => 5 + 2N - 1) 336 * `- double indirect node (5 + 2N) 337 * `- indirect node (6 + 2N) 338 * `- direct node 339 * ...... 340 * `- indirect node ((6 + 2N) + x(N + 1)) 341 * `- direct node 342 * ...... 343 * `- indirect node ((6 + 2N) + (N - 1)(N + 1)) 344 * `- direct node 345 */ 346 static inline bool IS_DNODE(struct page *node_page) 347 { 348 unsigned int ofs = ofs_of_node(node_page); 349 350 if (f2fs_has_xattr_block(ofs)) 351 return true; 352 353 if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK || 354 ofs == 5 + 2 * NIDS_PER_BLOCK) 355 return false; 356 if (ofs >= 6 + 2 * NIDS_PER_BLOCK) { 357 ofs -= 6 + 2 * NIDS_PER_BLOCK; 358 if (!((long int)ofs % (NIDS_PER_BLOCK + 1))) 359 return false; 360 } 361 return true; 362 } 363 364 static inline int set_nid(struct page *p, int off, nid_t nid, bool i) 365 { 366 struct f2fs_node *rn = F2FS_NODE(p); 367 368 f2fs_wait_on_page_writeback(p, NODE, true, true); 369 370 if (i) 371 rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid); 372 else 373 rn->in.nid[off] = cpu_to_le32(nid); 374 return set_page_dirty(p); 375 } 376 377 static inline nid_t get_nid(struct page *p, int off, bool i) 378 { 379 struct f2fs_node *rn = F2FS_NODE(p); 380 381 if (i) 382 return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]); 383 return le32_to_cpu(rn->in.nid[off]); 384 } 385 386 /* 387 * Coldness identification: 388 * - Mark cold files in f2fs_inode_info 389 * - Mark cold node blocks in their node footer 390 * - Mark cold data pages in page cache 391 */ 392 393 static inline int is_node(struct page *page, int type) 394 { 395 struct f2fs_node *rn = F2FS_NODE(page); 396 return le32_to_cpu(rn->footer.flag) & BIT(type); 397 } 398 399 #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT) 400 #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT) 401 #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT) 402 403 static inline void set_cold_node(struct page *page, bool is_dir) 404 { 405 struct f2fs_node *rn = F2FS_NODE(page); 406 unsigned int flag = le32_to_cpu(rn->footer.flag); 407 408 if (is_dir) 409 flag &= ~BIT(COLD_BIT_SHIFT); 410 else 411 flag |= BIT(COLD_BIT_SHIFT); 412 rn->footer.flag = cpu_to_le32(flag); 413 } 414 415 static inline void set_mark(struct page *page, int mark, int type) 416 { 417 struct f2fs_node *rn = F2FS_NODE(page); 418 unsigned int flag = le32_to_cpu(rn->footer.flag); 419 if (mark) 420 flag |= BIT(type); 421 else 422 flag &= ~BIT(type); 423 rn->footer.flag = cpu_to_le32(flag); 424 425 #ifdef CONFIG_F2FS_CHECK_FS 426 f2fs_inode_chksum_set(F2FS_P_SB(page), page); 427 #endif 428 } 429 #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT) 430 #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT) 431