| 1 |
|
| 2 | |
| 3 | * fs/f2fs/node.h
|
| 4 | *
|
| 5 | * Copyright (c) 2012 Samsung Electronics Co., Ltd.
|
| 6 | * http://www.samsung.com/
|
| 7 |
|
| 8 |
|
| 9 | #define START_NID(nid) (((nid) / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
|
| 10 |
|
| 11 |
|
| 12 | #define NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK)
|
| 13 |
|
| 14 |
|
| 15 | #define FREE_NID_PAGES 8
|
| 16 | #define MAX_FREE_NIDS (NAT_ENTRY_PER_BLOCK * FREE_NID_PAGES)
|
| 17 |
|
| 18 |
|
| 19 | #define SHRINK_NID_BATCH_SIZE 8
|
| 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 | IS_PREALLOC,
|
| 48 | };
|
| 49 |
|
| 50 | |
| 51 | * For node information
|
| 52 |
|
| 53 | struct node_info {
|
| 54 | nid_t nid;
|
| 55 | nid_t ino;
|
| 56 | block_t blk_addr;
|
| 57 | unsigned char version;
|
| 58 | unsigned char flag;
|
| 59 | };
|
| 60 |
|
| 61 | struct nat_entry {
|
| 62 | struct list_head list;
|
| 63 | struct node_info ni;
|
| 64 | };
|
| 65 |
|
| 66 | #define nat_get_nid(nat) ((nat)->ni.nid)
|
| 67 | #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
|
| 68 | #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
|
| 69 | #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
|
| 70 | #define nat_get_ino(nat) ((nat)->ni.ino)
|
| 71 | #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
|
| 72 | #define nat_get_version(nat) ((nat)->ni.version)
|
| 73 | #define nat_set_version(nat, v) ((nat)->ni.version = (v))
|
| 74 |
|
| 75 | #define inc_node_version(version) (++(version))
|
| 76 |
|
| 77 | static inline void copy_node_info(struct node_info *dst,
|
| 78 | struct node_info *src)
|
| 79 | {
|
| 80 | dst->nid = src->nid;
|
| 81 | dst->ino = src->ino;
|
| 82 | dst->blk_addr = src->blk_addr;
|
| 83 | dst->version = src->version;
|
| 84 |
|
| 85 | }
|
| 86 |
|
| 87 | static inline void set_nat_flag(struct nat_entry *ne,
|
| 88 | unsigned int type, bool set)
|
| 89 | {
|
| 90 | unsigned char mask = 0x01 << type;
|
| 91 | if (set)
|
| 92 | ne->ni.flag |= mask;
|
| 93 | else
|
| 94 | ne->ni.flag &= ~mask;
|
| 95 | }
|
| 96 |
|
| 97 | static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
|
| 98 | {
|
| 99 | unsigned char mask = 0x01 << type;
|
| 100 | return ne->ni.flag & mask;
|
| 101 | }
|
| 102 |
|
| 103 | static inline void nat_reset_flag(struct nat_entry *ne)
|
| 104 | {
|
| 105 |
|
| 106 | set_nat_flag(ne, IS_CHECKPOINTED, true);
|
| 107 | set_nat_flag(ne, HAS_FSYNCED_INODE, false);
|
| 108 | set_nat_flag(ne, HAS_LAST_FSYNC, true);
|
| 109 | }
|
| 110 |
|
| 111 | static inline void node_info_from_raw_nat(struct node_info *ni,
|
| 112 | struct f2fs_nat_entry *raw_ne)
|
| 113 | {
|
| 114 | ni->ino = le32_to_cpu(raw_ne->ino);
|
| 115 | ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
|
| 116 | ni->version = raw_ne->version;
|
| 117 | }
|
| 118 |
|
| 119 | static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
|
| 120 | struct node_info *ni)
|
| 121 | {
|
| 122 | raw_ne->ino = cpu_to_le32(ni->ino);
|
| 123 | raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
|
| 124 | raw_ne->version = ni->version;
|
| 125 | }
|
| 126 |
|
| 127 | static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
|
| 128 | {
|
| 129 | return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid *
|
| 130 | NM_I(sbi)->dirty_nats_ratio / 100;
|
| 131 | }
|
| 132 |
|
| 133 | static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
|
| 134 | {
|
| 135 | return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD;
|
| 136 | }
|
| 137 |
|
| 138 | static inline bool excess_dirty_nodes(struct f2fs_sb_info *sbi)
|
| 139 | {
|
| 140 | return get_pages(sbi, F2FS_DIRTY_NODES) >= sbi->blocks_per_seg * 8;
|
| 141 | }
|
| 142 |
|
| 143 | enum mem_type {
|
| 144 | FREE_NIDS,
|
| 145 | NAT_ENTRIES,
|
| 146 | DIRTY_DENTS,
|
| 147 | INO_ENTRIES,
|
| 148 | EXTENT_CACHE,
|
| 149 | INMEM_PAGES,
|
| 150 | DISCARD_CACHE,
|
| 151 | BASE_CHECK,
|
| 152 | };
|
| 153 |
|
| 154 | struct nat_entry_set {
|
| 155 | struct list_head set_list;
|
| 156 | struct list_head entry_list;
|
| 157 | nid_t set;
|
| 158 | unsigned int entry_cnt;
|
| 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] <= 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 ^= 1 << 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 |
|
| 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 |
|
| 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 | static inline int is_cold_data(struct page *page)
|
| 393 | {
|
| 394 | return PageChecked(page);
|
| 395 | }
|
| 396 |
|
| 397 | static inline void set_cold_data(struct page *page)
|
| 398 | {
|
| 399 | SetPageChecked(page);
|
| 400 | }
|
| 401 |
|
| 402 | static inline void clear_cold_data(struct page *page)
|
| 403 | {
|
| 404 | ClearPageChecked(page);
|
| 405 | }
|
| 406 |
|
| 407 | static inline int is_node(struct page *page, int type)
|
| 408 | {
|
| 409 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 410 | return le32_to_cpu(rn->footer.flag) & (1 << type);
|
| 411 | }
|
| 412 |
|
| 413 | #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
|
| 414 | #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
|
| 415 | #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
|
| 416 |
|
| 417 | static inline int is_inline_node(struct page *page)
|
| 418 | {
|
| 419 | return PageChecked(page);
|
| 420 | }
|
| 421 |
|
| 422 | static inline void set_inline_node(struct page *page)
|
| 423 | {
|
| 424 | SetPageChecked(page);
|
| 425 | }
|
| 426 |
|
| 427 | static inline void clear_inline_node(struct page *page)
|
| 428 | {
|
| 429 | ClearPageChecked(page);
|
| 430 | }
|
| 431 |
|
| 432 | static inline void set_cold_node(struct page *page, bool is_dir)
|
| 433 | {
|
| 434 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 435 | unsigned int flag = le32_to_cpu(rn->footer.flag);
|
| 436 |
|
| 437 | if (is_dir)
|
| 438 | flag &= ~(0x1 << COLD_BIT_SHIFT);
|
| 439 | else
|
| 440 | flag |= (0x1 << COLD_BIT_SHIFT);
|
| 441 | rn->footer.flag = cpu_to_le32(flag);
|
| 442 | }
|
| 443 |
|
| 444 | static inline void set_mark(struct page *page, int mark, int type)
|
| 445 | {
|
| 446 | struct f2fs_node *rn = F2FS_NODE(page);
|
| 447 | unsigned int flag = le32_to_cpu(rn->footer.flag);
|
| 448 | if (mark)
|
| 449 | flag |= (0x1 << type);
|
| 450 | else
|
| 451 | flag &= ~(0x1 << type);
|
| 452 | rn->footer.flag = cpu_to_le32(flag);
|
| 453 |
|
| 454 | #ifdef CONFIG_F2FS_CHECK_FS
|
| 455 | f2fs_inode_chksum_set(F2FS_P_SB(page), page);
|
| 456 | #endif
|
| 457 | }
|
| 458 | #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
|
| 459 | #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)
|