| 1 | |
| 2 | * fs/f2fs/node.h
|
| 3 | *
|
| 4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd.
|
| 5 | * http://www.samsung.com/
|
| 6 | *
|
| 7 | * This program is free software; you can redistribute it and/or modify
|
| 8 | * it under the terms of the GNU General Public License version 2 as
|
| 9 | * published by the Free Software Foundation.
|
| 10 |
|
| 11 |
|
| 12 | #define START_NID(nid) (((nid) / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
|
| 13 |
|
| 14 |
|
| 15 | #define NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK)
|
| 16 |
|
| 17 |
|
| 18 | #define FREE_NID_PAGES 8
|
| 19 | #define MAX_FREE_NIDS (NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES)
|
| 20 |
|
| 21 | #define DEF_RA_NID_PAGES 0
|
| 22 |
|
| 23 |
|
| 24 | #define MAX_RA_NODE 128
|
| 25 |
|
| 26 |
|
| 27 | #define DEF_RAM_THRESHOLD 1
|
| 28 |
|
| 29 |
|
| 30 | #define DEF_DIRTY_NAT_RATIO_THRESHOLD 10
|
| 31 |
|
| 32 | #define DEF_NAT_CACHE_THRESHOLD 100000
|
| 33 |
|
| 34 |
|
| 35 | #define NATVEC_SIZE 64
|
| 36 | #define SETVEC_SIZE 32
|
| 37 |
|
| 38 |
|
| 39 | #define LOCKED_PAGE 1
|
| 40 |
|
| 41 |
|
| 42 | enum {
|
| 43 | IS_CHECKPOINTED,
|
| 44 | HAS_FSYNCED_INODE,
|
| 45 | HAS_LAST_FSYNC,
|
| 46 | IS_DIRTY,
|
| 47 | };
|
| 48 |
|
| 49 | |
| 50 | * For node information
|
| 51 |
|
| 52 | struct node_info {
|
| 53 | nid_t nid;
|
| 54 | nid_t ino;
|
| 55 | block_t blk_addr;
|
| 56 | unsigned char version;
|
| 57 | unsigned char flag;
|
| 58 | };
|
| 59 |
|
| 60 | struct nat_entry {
|
| 61 | struct list_head list;
|
| 62 | struct node_info ni;
|
| 63 | };
|
| 64 |
|
| 65 | #define nat_get_nid(nat) ((nat)->ni.nid)
|
| 66 | #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
|
| 67 | #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
|
| 68 | #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
|
| 69 | #define nat_get_ino(nat) ((nat)->ni.ino)
|
| 70 | #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
|
| 71 | #define nat_get_version(nat) ((nat)->ni.version)
|
| 72 | #define nat_set_version(nat, v) ((nat)->ni.version = (v))
|
| 73 |
|
| 74 | #define inc_node_version(version) (++(version))
|
| 75 |
|
| 76 | static inline void copy_node_info(struct node_info *dst,
|
| 77 | struct node_info *src)
|
| 78 | {
|
| 79 | dst->nid = src->nid;
|
| 80 | dst->ino = src->ino;
|
| 81 | dst->blk_addr = src->blk_addr;
|
| 82 | dst->version = src->version;
|
| 83 |
|
| 84 | }
|
| 85 |
|
| 86 | static inline void set_nat_flag(struct nat_entry *ne,
|
| 87 | unsigned int type, bool set)
|
| 88 | {
|
| 89 | unsigned char mask = 0x01 << type;
|
| 90 | if (set)
|
| 91 | ne->ni.flag |= mask;
|
| 92 | else
|
| 93 | ne->ni.flag &= ~mask;
|
| 94 | }
|
| 95 |
|
| 96 | static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
|
| 97 | {
|
| 98 | unsigned char mask = 0x01 << type;
|
| 99 | return ne->ni.flag & mask;
|
| 100 | }
|
| 101 |
|
| 102 | static inline void nat_reset_flag(struct nat_entry *ne)
|
| 103 | {
|
| 104 |
|
| 105 | set_nat_flag(ne, IS_CHECKPOINTED, true);
|
| 106 | set_nat_flag(ne, HAS_FSYNCED_INODE, false);
|
| 107 | set_nat_flag(ne, HAS_LAST_FSYNC, true);
|
| 108 | }
|
| 109 |
|
| 110 | static inline void node_info_from_raw_nat(struct node_info *ni,
|
| 111 | struct f2fs_nat_entry *raw_ne)
|
| 112 | {
|
| 113 | ni->ino = le32_to_cpu(raw_ne->ino);
|
| 114 | ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
|
| 115 | ni->version = raw_ne->version;
|
| 116 | }
|
| 117 |
|
| 118 | static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
|
| 119 | struct node_info *ni)
|
| 120 | {
|
| 121 | raw_ne->ino = cpu_to_le32(ni->ino);
|
| 122 | raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
|
| 123 | raw_ne->version = ni->version;
|
| 124 | }
|
| 125 |
|
| 126 | static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
|
| 127 | {
|
| 128 | return NM_I(sbi)->dirty_nat_cnt >= NM_I(sbi)->max_nid *
|
| 129 | NM_I(sbi)->dirty_nats_ratio / 100;
|
| 130 | }
|
| 131 |
|
| 132 | static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
|
| 133 | {
|
| 134 | return NM_I(sbi)->nat_cnt >= DEF_NAT_CACHE_THRESHOLD;
|
| 135 | }
|
| 136 |
|
| 137 | enum mem_type {
|
| 138 | FREE_NIDS,
|
| 139 | NAT_ENTRIES,
|
| 140 | DIRTY_DENTS,
|
| 141 | INO_ENTRIES,
|
| 142 | EXTENT_CACHE,
|
| 143 | BASE_CHECK,
|
| 144 | };
|
| 145 |
|
| 146 | struct nat_entry_set {
|
| 147 | struct list_head set_list;
|
| 148 | struct list_head entry_list;
|
| 149 | nid_t set;
|
| 150 | unsigned int entry_cnt;
|
| 151 | };
|
| 152 |
|
| 153 | |
| 154 | * For free nid mangement
|
| 155 |
|
| 156 | enum nid_state {
|
| 157 | NID_NEW,
|
| 158 | NID_ALLOC
|
| 159 | };
|
| 160 |
|
| 161 | struct free_nid {
|
| 162 | struct list_head list;
|
| 163 | nid_t nid;
|
| 164 | int state;
|
| 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_LIST] <= 0) {
|
| 174 | spin_unlock(&nm_i->nid_list_lock);
|
| 175 | return;
|
| 176 | }
|
| 177 | fnid = list_first_entry(&nm_i->nid_list[FREE_NID_LIST],
|
| 178 | struct free_nid, list);
|
| 179 | *nid = fnid->nid;
|
| 180 | spin_unlock(&nm_i->nid_list_lock);
|
| 181 | }
|
| 182 |
|
| 183 | |
| 184 | * inline functions
|
| 185 |
|
| 186 | static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
|
| 187 | {
|
| 188 | struct f2fs_nm_info *nm_i = NM_I(sbi);
|
| 189 |
|
| 190 | #ifdef CONFIG_F2FS_CHECK_FS
|
| 191 | if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
|
| 192 | nm_i->bitmap_size))
|
| 193 | f2fs_bug_on(sbi, 1);
|
| 194 | #endif
|
| 195 | memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
|
| 196 | }
|
| 197 |
|
| 198 | static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
|
| 199 | {
|
| 200 | struct f2fs_nm_info *nm_i = NM_I(sbi);
|
| 201 | pgoff_t block_off;
|
| 202 | pgoff_t block_addr;
|
| 203 |
|
| 204 | |
| 205 | * block_off = segment_off * 512 + off_in_segment
|
| 206 | * OLD = (segment_off * 512) * 2 + off_in_segment
|
| 207 | * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
|
| 208 |
|
| 209 | block_off = NAT_BLOCK_OFFSET(start);
|
| 210 |
|
| 211 | block_addr = (pgoff_t)(nm_i->nat_blkaddr +
|
| 212 | (block_off << 1) -
|
| 213 | (block_off & (sbi->blocks_per_seg - 1)));
|
| 214 |
|
| 215 | if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
|
| 216 | block_addr += sbi->blocks_per_seg;
|
| 217 |
|
| 218 | return block_addr;
|
| 219 | }
|
| 220 |
|
| 221 | static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
|
| 222 | pgoff_t block_addr)
|
| 223 | {
|
| 224 | struct f2fs_nm_info *nm_i = NM_I(sbi);
|
| 225 |
|
| 226 | block_addr -= nm_i->nat_blkaddr;
|
| 227 | block_addr ^= 1 << sbi->log_blocks_per_seg;
|
| 228 | return block_addr + nm_i->nat_blkaddr;
|
| 229 | }
|
| 230 |
|
| 231 | static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
|
| 232 | {
|
| 233 | unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
|
| 234 |
|
| 235 | f2fs_change_bit(block_off, nm_i->nat_bitmap);
|
| 236 | #ifdef CONFIG_F2FS_CHECK_FS
|
| 237 | f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
|
| 238 | #endif
|
| 239 | }
|
| 240 |
|
| 241 | static inline nid_t ino_of_node(struct page *node_page)
|
| 242 | {
|
| 243 | struct f2fs_node *rn = F2FS_NODE(node_page);
|
| 244 | return le32_to_cpu(rn->footer.ino);
|
| 245 | }
|
| 246 |
|
| 247 | static inline nid_t nid_of_node(struct page *node_page)
|
| 248 | {
|
| 249 | struct f2fs_node *rn = F2FS_NODE(node_page);
|
| 250 | return le32_to_cpu(rn->footer.nid);
|
| 251 | }
|
| 252 |
|
| 253 | static inline unsigned int ofs_of_node(struct page *node_page)
|
| 254 | {
|
| 255 | struct f2fs_node *rn = F2FS_NODE(node_page);
|
| 256 | unsigned flag = le32_to_cpu(rn->footer.flag);
|
| 257 | return flag >> OFFSET_BIT_SHIFT;
|
| 258 | }
|
| 259 |
|
| 260 | static inline __u64 cpver_of_node(struct page *node_page)
|
| 261 | {
|
| 262 | struct f2fs_node *rn = F2FS_NODE(node_page);
|
| 263 | return le64_to_cpu(rn->footer.cp_ver);
|
| 264 | }
|
| 265 |
|
| 266 | static inline block_t next_blkaddr_of_node(struct page *node_page)
|
| 267 | {
|
| 268 | struct f2fs_node *rn = F2FS_NODE(node_page);
|
| 269 | return le32_to_cpu(rn->footer.next_blkaddr);
|
| 270 | }
|
| 271 |
|
| 272 | static inline void fill_node_footer(struct page *page, nid_t nid,
|
| 273 | nid_t ino, unsigned int ofs, bool reset)
|
| 274 | {
|
| 275 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 276 | unsigned int old_flag = 0;
|
| 277 |
|
| 278 | if (reset)
|
| 279 | memset(rn, 0, sizeof(*rn));
|
| 280 | else
|
| 281 | old_flag = le32_to_cpu(rn->footer.flag);
|
| 282 |
|
| 283 | rn->footer.nid = cpu_to_le32(nid);
|
| 284 | rn->footer.ino = cpu_to_le32(ino);
|
| 285 |
|
| 286 |
|
| 287 | rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
|
| 288 | (old_flag & OFFSET_BIT_MASK));
|
| 289 | }
|
| 290 |
|
| 291 | static inline void copy_node_footer(struct page *dst, struct page *src)
|
| 292 | {
|
| 293 | struct f2fs_node *src_rn = F2FS_NODE(src);
|
| 294 | struct f2fs_node *dst_rn = F2FS_NODE(dst);
|
| 295 | memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
|
| 296 | }
|
| 297 |
|
| 298 | static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
|
| 299 | {
|
| 300 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
|
| 301 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 302 | __u64 cp_ver = cur_cp_version(ckpt);
|
| 303 |
|
| 304 | if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
|
| 305 | cp_ver |= (cur_cp_crc(ckpt) << 32);
|
| 306 |
|
| 307 | rn->footer.cp_ver = cpu_to_le64(cp_ver);
|
| 308 | rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
|
| 309 | }
|
| 310 |
|
| 311 | static inline bool is_recoverable_dnode(struct page *page)
|
| 312 | {
|
| 313 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
|
| 314 | __u64 cp_ver = cur_cp_version(ckpt);
|
| 315 |
|
| 316 | if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
|
| 317 | cp_ver |= (cur_cp_crc(ckpt) << 32);
|
| 318 |
|
| 319 | return cp_ver == cpver_of_node(page);
|
| 320 | }
|
| 321 |
|
| 322 | |
| 323 | * f2fs assigns the following node offsets described as (num).
|
| 324 | * N = NIDS_PER_BLOCK
|
| 325 | *
|
| 326 | * Inode block (0)
|
| 327 | * |- direct node (1)
|
| 328 | * |- direct node (2)
|
| 329 | * |- indirect node (3)
|
| 330 | * | `- direct node (4 => 4 + N - 1)
|
| 331 | * |- indirect node (4 + N)
|
| 332 | * | `- direct node (5 + N => 5 + 2N - 1)
|
| 333 | * `- double indirect node (5 + 2N)
|
| 334 | * `- indirect node (6 + 2N)
|
| 335 | * `- direct node
|
| 336 | * ......
|
| 337 | * `- indirect node ((6 + 2N) + x(N + 1))
|
| 338 | * `- direct node
|
| 339 | * ......
|
| 340 | * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
|
| 341 | * `- direct node
|
| 342 |
|
| 343 | static inline bool IS_DNODE(struct page *node_page)
|
| 344 | {
|
| 345 | unsigned int ofs = ofs_of_node(node_page);
|
| 346 |
|
| 347 | if (f2fs_has_xattr_block(ofs))
|
| 348 | return true;
|
| 349 |
|
| 350 | if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
|
| 351 | ofs == 5 + 2 * NIDS_PER_BLOCK)
|
| 352 | return false;
|
| 353 | if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
|
| 354 | ofs -= 6 + 2 * NIDS_PER_BLOCK;
|
| 355 | if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
|
| 356 | return false;
|
| 357 | }
|
| 358 | return true;
|
| 359 | }
|
| 360 |
|
| 361 | static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
|
| 362 | {
|
| 363 | struct f2fs_node *rn = F2FS_NODE(p);
|
| 364 |
|
| 365 | f2fs_wait_on_page_writeback(p, NODE, true);
|
| 366 |
|
| 367 | if (i)
|
| 368 | rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
|
| 369 | else
|
| 370 | rn->in.nid[off] = cpu_to_le32(nid);
|
| 371 | return set_page_dirty(p);
|
| 372 | }
|
| 373 |
|
| 374 | static inline nid_t get_nid(struct page *p, int off, bool i)
|
| 375 | {
|
| 376 | struct f2fs_node *rn = F2FS_NODE(p);
|
| 377 |
|
| 378 | if (i)
|
| 379 | return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
|
| 380 | return le32_to_cpu(rn->in.nid[off]);
|
| 381 | }
|
| 382 |
|
| 383 | |
| 384 | * Coldness identification:
|
| 385 | * - Mark cold files in f2fs_inode_info
|
| 386 | * - Mark cold node blocks in their node footer
|
| 387 | * - Mark cold data pages in page cache
|
| 388 |
|
| 389 | static inline int is_cold_data(struct page *page)
|
| 390 | {
|
| 391 | return PageChecked(page);
|
| 392 | }
|
| 393 |
|
| 394 | static inline void set_cold_data(struct page *page)
|
| 395 | {
|
| 396 | SetPageChecked(page);
|
| 397 | }
|
| 398 |
|
| 399 | static inline void clear_cold_data(struct page *page)
|
| 400 | {
|
| 401 | ClearPageChecked(page);
|
| 402 | }
|
| 403 |
|
| 404 | static inline int is_node(struct page *page, int type)
|
| 405 | {
|
| 406 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 407 | return le32_to_cpu(rn->footer.flag) & (1 << type);
|
| 408 | }
|
| 409 |
|
| 410 | #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
|
| 411 | #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
|
| 412 | #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
|
| 413 |
|
| 414 | static inline int is_inline_node(struct page *page)
|
| 415 | {
|
| 416 | return PageChecked(page);
|
| 417 | }
|
| 418 |
|
| 419 | static inline void set_inline_node(struct page *page)
|
| 420 | {
|
| 421 | SetPageChecked(page);
|
| 422 | }
|
| 423 |
|
| 424 | static inline void clear_inline_node(struct page *page)
|
| 425 | {
|
| 426 | ClearPageChecked(page);
|
| 427 | }
|
| 428 |
|
| 429 | static inline void set_cold_node(struct inode *inode, struct page *page)
|
| 430 | {
|
| 431 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 432 | unsigned int flag = le32_to_cpu(rn->footer.flag);
|
| 433 |
|
| 434 | if (S_ISDIR(inode->i_mode))
|
| 435 | flag &= ~(0x1 << COLD_BIT_SHIFT);
|
| 436 | else
|
| 437 | flag |= (0x1 << COLD_BIT_SHIFT);
|
| 438 | rn->footer.flag = cpu_to_le32(flag);
|
| 439 | }
|
| 440 |
|
| 441 | static inline void set_mark(struct page *page, int mark, int type)
|
| 442 | {
|
| 443 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 444 | unsigned int flag = le32_to_cpu(rn->footer.flag);
|
| 445 | if (mark)
|
| 446 | flag |= (0x1 << type);
|
| 447 | else
|
| 448 | flag &= ~(0x1 << type);
|
| 449 | rn->footer.flag = cpu_to_le32(flag);
|
| 450 | }
|
| 451 | #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
|
| 452 | #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)
|