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 NAT_VEC_SIZE 32 |
39 | |
40 | /* return value for read_node_page */ |
41 | #define LOCKED_PAGE 1 |
42 | |
43 | /* check pinned file's alignment status of physical blocks */ |
44 | #define FILE_NOT_ALIGNED 1 |
45 | |
46 | /* For flag in struct node_info */ |
47 | enum { |
48 | IS_CHECKPOINTED, /* is it checkpointed before? */ |
49 | HAS_FSYNCED_INODE, /* is the inode fsynced before? */ |
50 | HAS_LAST_FSYNC, /* has the latest node fsync mark? */ |
51 | IS_DIRTY, /* this nat entry is dirty? */ |
52 | IS_PREALLOC, /* nat entry is preallocated */ |
53 | }; |
54 | |
55 | /* For node type in __get_node_folio() */ |
56 | enum node_type { |
57 | NODE_TYPE_REGULAR, |
58 | NODE_TYPE_INODE, |
59 | NODE_TYPE_XATTR, |
60 | }; |
61 | |
62 | /* |
63 | * For node information |
64 | */ |
65 | struct node_info { |
66 | nid_t nid; /* node id */ |
67 | nid_t ino; /* inode number of the node's owner */ |
68 | block_t blk_addr; /* block address of the node */ |
69 | unsigned char version; /* version of the node */ |
70 | unsigned char flag; /* for node information bits */ |
71 | }; |
72 | |
73 | struct nat_entry { |
74 | struct list_head list; /* for clean or dirty nat list */ |
75 | struct node_info ni; /* in-memory node information */ |
76 | }; |
77 | |
78 | #define nat_get_nid(nat) ((nat)->ni.nid) |
79 | #define nat_set_nid(nat, n) ((nat)->ni.nid = (n)) |
80 | #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr) |
81 | #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b)) |
82 | #define nat_get_ino(nat) ((nat)->ni.ino) |
83 | #define nat_set_ino(nat, i) ((nat)->ni.ino = (i)) |
84 | #define nat_get_version(nat) ((nat)->ni.version) |
85 | #define nat_set_version(nat, v) ((nat)->ni.version = (v)) |
86 | |
87 | #define inc_node_version(version) (++(version)) |
88 | |
89 | static inline void copy_node_info(struct node_info *dst, |
90 | struct node_info *src) |
91 | { |
92 | dst->nid = src->nid; |
93 | dst->ino = src->ino; |
94 | dst->blk_addr = src->blk_addr; |
95 | dst->version = src->version; |
96 | /* should not copy flag here */ |
97 | } |
98 | |
99 | static inline void set_nat_flag(struct nat_entry *ne, |
100 | unsigned int type, bool set) |
101 | { |
102 | if (set) |
103 | ne->ni.flag |= BIT(type); |
104 | else |
105 | ne->ni.flag &= ~BIT(type); |
106 | } |
107 | |
108 | static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type) |
109 | { |
110 | return ne->ni.flag & BIT(type); |
111 | } |
112 | |
113 | static inline void nat_reset_flag(struct nat_entry *ne) |
114 | { |
115 | /* these states can be set only after checkpoint was done */ |
116 | set_nat_flag(ne, type: IS_CHECKPOINTED, set: true); |
117 | set_nat_flag(ne, type: HAS_FSYNCED_INODE, set: false); |
118 | set_nat_flag(ne, type: HAS_LAST_FSYNC, set: true); |
119 | } |
120 | |
121 | static inline void node_info_from_raw_nat(struct node_info *ni, |
122 | struct f2fs_nat_entry *raw_ne) |
123 | { |
124 | ni->ino = le32_to_cpu(raw_ne->ino); |
125 | ni->blk_addr = le32_to_cpu(raw_ne->block_addr); |
126 | ni->version = raw_ne->version; |
127 | } |
128 | |
129 | static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne, |
130 | struct node_info *ni) |
131 | { |
132 | raw_ne->ino = cpu_to_le32(ni->ino); |
133 | raw_ne->block_addr = cpu_to_le32(ni->blk_addr); |
134 | raw_ne->version = ni->version; |
135 | } |
136 | |
137 | static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi) |
138 | { |
139 | return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid * |
140 | NM_I(sbi)->dirty_nats_ratio / 100; |
141 | } |
142 | |
143 | static inline bool excess_cached_nats(struct f2fs_sb_info *sbi) |
144 | { |
145 | return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD; |
146 | } |
147 | |
148 | enum mem_type { |
149 | FREE_NIDS, /* indicates the free nid list */ |
150 | NAT_ENTRIES, /* indicates the cached nat entry */ |
151 | DIRTY_DENTS, /* indicates dirty dentry pages */ |
152 | INO_ENTRIES, /* indicates inode entries */ |
153 | READ_EXTENT_CACHE, /* indicates read extent cache */ |
154 | AGE_EXTENT_CACHE, /* indicates age extent cache */ |
155 | DISCARD_CACHE, /* indicates memory of cached discard cmds */ |
156 | COMPRESS_PAGE, /* indicates memory of cached compressed pages */ |
157 | BASE_CHECK, /* check kernel status */ |
158 | }; |
159 | |
160 | struct nat_entry_set { |
161 | struct list_head set_list; /* link with other nat sets */ |
162 | struct list_head entry_list; /* link with dirty nat entries */ |
163 | nid_t set; /* set number*/ |
164 | unsigned int entry_cnt; /* the # of nat entries in set */ |
165 | }; |
166 | |
167 | struct free_nid { |
168 | struct list_head list; /* for free node id list */ |
169 | nid_t nid; /* node id */ |
170 | int state; /* in use or not: FREE_NID or PREALLOC_NID */ |
171 | }; |
172 | |
173 | static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid) |
174 | { |
175 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
176 | struct free_nid *fnid; |
177 | |
178 | spin_lock(lock: &nm_i->nid_list_lock); |
179 | if (nm_i->nid_cnt[FREE_NID] <= 0) { |
180 | spin_unlock(lock: &nm_i->nid_list_lock); |
181 | return; |
182 | } |
183 | fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list); |
184 | *nid = fnid->nid; |
185 | spin_unlock(lock: &nm_i->nid_list_lock); |
186 | } |
187 | |
188 | /* |
189 | * inline functions |
190 | */ |
191 | static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr) |
192 | { |
193 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
194 | |
195 | #ifdef CONFIG_F2FS_CHECK_FS |
196 | if (memcmp(p: nm_i->nat_bitmap, q: nm_i->nat_bitmap_mir, |
197 | size: nm_i->bitmap_size)) |
198 | f2fs_bug_on(sbi, 1); |
199 | #endif |
200 | memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size); |
201 | } |
202 | |
203 | static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start) |
204 | { |
205 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
206 | pgoff_t block_off; |
207 | pgoff_t block_addr; |
208 | |
209 | /* |
210 | * block_off = segment_off * 512 + off_in_segment |
211 | * OLD = (segment_off * 512) * 2 + off_in_segment |
212 | * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment |
213 | */ |
214 | block_off = NAT_BLOCK_OFFSET(start); |
215 | |
216 | block_addr = (pgoff_t)(nm_i->nat_blkaddr + |
217 | (block_off << 1) - |
218 | (block_off & (BLKS_PER_SEG(sbi) - 1))); |
219 | |
220 | if (f2fs_test_bit(nr: block_off, addr: nm_i->nat_bitmap)) |
221 | block_addr += BLKS_PER_SEG(sbi); |
222 | |
223 | return block_addr; |
224 | } |
225 | |
226 | static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi, |
227 | pgoff_t block_addr) |
228 | { |
229 | struct f2fs_nm_info *nm_i = NM_I(sbi); |
230 | |
231 | block_addr -= nm_i->nat_blkaddr; |
232 | block_addr ^= BIT(sbi->log_blocks_per_seg); |
233 | return block_addr + nm_i->nat_blkaddr; |
234 | } |
235 | |
236 | static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid) |
237 | { |
238 | unsigned int block_off = NAT_BLOCK_OFFSET(start_nid); |
239 | |
240 | f2fs_change_bit(nr: block_off, addr: nm_i->nat_bitmap); |
241 | #ifdef CONFIG_F2FS_CHECK_FS |
242 | f2fs_change_bit(nr: block_off, addr: nm_i->nat_bitmap_mir); |
243 | #endif |
244 | } |
245 | |
246 | static inline nid_t ino_of_node(struct page *node_page) |
247 | { |
248 | struct f2fs_node *rn = F2FS_NODE(page: node_page); |
249 | return le32_to_cpu(rn->footer.ino); |
250 | } |
251 | |
252 | static inline nid_t nid_of_node(struct page *node_page) |
253 | { |
254 | struct f2fs_node *rn = F2FS_NODE(page: node_page); |
255 | return le32_to_cpu(rn->footer.nid); |
256 | } |
257 | |
258 | static inline unsigned int ofs_of_node(const struct page *node_page) |
259 | { |
260 | struct f2fs_node *rn = F2FS_NODE(page: node_page); |
261 | unsigned flag = le32_to_cpu(rn->footer.flag); |
262 | return flag >> OFFSET_BIT_SHIFT; |
263 | } |
264 | |
265 | static inline __u64 cpver_of_node(struct page *node_page) |
266 | { |
267 | struct f2fs_node *rn = F2FS_NODE(page: node_page); |
268 | return le64_to_cpu(rn->footer.cp_ver); |
269 | } |
270 | |
271 | static inline block_t next_blkaddr_of_node(struct folio *node_folio) |
272 | { |
273 | struct f2fs_node *rn = F2FS_NODE(page: &node_folio->page); |
274 | return le32_to_cpu(rn->footer.next_blkaddr); |
275 | } |
276 | |
277 | static inline void (struct page *page, nid_t nid, |
278 | nid_t ino, unsigned int ofs, bool reset) |
279 | { |
280 | struct f2fs_node *rn = F2FS_NODE(page); |
281 | unsigned int old_flag = 0; |
282 | |
283 | if (reset) |
284 | memset(rn, 0, sizeof(*rn)); |
285 | else |
286 | old_flag = le32_to_cpu(rn->footer.flag); |
287 | |
288 | rn->footer.nid = cpu_to_le32(nid); |
289 | rn->footer.ino = cpu_to_le32(ino); |
290 | |
291 | /* should remain old flag bits such as COLD_BIT_SHIFT */ |
292 | rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) | |
293 | (old_flag & OFFSET_BIT_MASK)); |
294 | } |
295 | |
296 | static inline void (struct page *dst, struct page *src) |
297 | { |
298 | struct f2fs_node *src_rn = F2FS_NODE(page: src); |
299 | struct f2fs_node *dst_rn = F2FS_NODE(page: dst); |
300 | memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer)); |
301 | } |
302 | |
303 | static inline void (struct page *page, block_t blkaddr) |
304 | { |
305 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi: F2FS_P_SB(page)); |
306 | struct f2fs_node *rn = F2FS_NODE(page); |
307 | __u64 cp_ver = cur_cp_version(cp: ckpt); |
308 | |
309 | if (__is_set_ckpt_flags(cp: ckpt, CP_CRC_RECOVERY_FLAG)) |
310 | cp_ver |= (cur_cp_crc(cp: ckpt) << 32); |
311 | |
312 | rn->footer.cp_ver = cpu_to_le64(cp_ver); |
313 | rn->footer.next_blkaddr = cpu_to_le32(blkaddr); |
314 | } |
315 | |
316 | static inline bool is_recoverable_dnode(struct page *page) |
317 | { |
318 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi: F2FS_P_SB(page)); |
319 | __u64 cp_ver = cur_cp_version(cp: ckpt); |
320 | |
321 | /* Don't care crc part, if fsck.f2fs sets it. */ |
322 | if (__is_set_ckpt_flags(cp: ckpt, CP_NOCRC_RECOVERY_FLAG)) |
323 | return (cp_ver << 32) == (cpver_of_node(node_page: page) << 32); |
324 | |
325 | if (__is_set_ckpt_flags(cp: ckpt, CP_CRC_RECOVERY_FLAG)) |
326 | cp_ver |= (cur_cp_crc(cp: ckpt) << 32); |
327 | |
328 | return cp_ver == cpver_of_node(node_page: page); |
329 | } |
330 | |
331 | /* |
332 | * f2fs assigns the following node offsets described as (num). |
333 | * N = NIDS_PER_BLOCK |
334 | * |
335 | * Inode block (0) |
336 | * |- direct node (1) |
337 | * |- direct node (2) |
338 | * |- indirect node (3) |
339 | * | `- direct node (4 => 4 + N - 1) |
340 | * |- indirect node (4 + N) |
341 | * | `- direct node (5 + N => 5 + 2N - 1) |
342 | * `- double indirect node (5 + 2N) |
343 | * `- indirect node (6 + 2N) |
344 | * `- direct node |
345 | * ...... |
346 | * `- indirect node ((6 + 2N) + x(N + 1)) |
347 | * `- direct node |
348 | * ...... |
349 | * `- indirect node ((6 + 2N) + (N - 1)(N + 1)) |
350 | * `- direct node |
351 | */ |
352 | static inline bool IS_DNODE(const struct page *node_page) |
353 | { |
354 | unsigned int ofs = ofs_of_node(node_page); |
355 | |
356 | if (f2fs_has_xattr_block(ofs)) |
357 | return true; |
358 | |
359 | if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK || |
360 | ofs == 5 + 2 * NIDS_PER_BLOCK) |
361 | return false; |
362 | if (ofs >= 6 + 2 * NIDS_PER_BLOCK) { |
363 | ofs -= 6 + 2 * NIDS_PER_BLOCK; |
364 | if (!((long int)ofs % (NIDS_PER_BLOCK + 1))) |
365 | return false; |
366 | } |
367 | return true; |
368 | } |
369 | |
370 | static inline int set_nid(struct folio *folio, int off, nid_t nid, bool i) |
371 | { |
372 | struct f2fs_node *rn = F2FS_NODE(page: &folio->page); |
373 | |
374 | f2fs_folio_wait_writeback(folio, type: NODE, ordered: true, locked: true); |
375 | |
376 | if (i) |
377 | rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid); |
378 | else |
379 | rn->in.nid[off] = cpu_to_le32(nid); |
380 | return folio_mark_dirty(folio); |
381 | } |
382 | |
383 | static inline nid_t get_nid(struct page *p, int off, bool i) |
384 | { |
385 | struct f2fs_node *rn = F2FS_NODE(page: p); |
386 | |
387 | if (i) |
388 | return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]); |
389 | return le32_to_cpu(rn->in.nid[off]); |
390 | } |
391 | |
392 | /* |
393 | * Coldness identification: |
394 | * - Mark cold files in f2fs_inode_info |
395 | * - Mark cold node blocks in their node footer |
396 | * - Mark cold data pages in page cache |
397 | */ |
398 | |
399 | static inline int is_node(const struct page *page, int type) |
400 | { |
401 | struct f2fs_node *rn = F2FS_NODE(page); |
402 | return le32_to_cpu(rn->footer.flag) & BIT(type); |
403 | } |
404 | |
405 | #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT) |
406 | #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT) |
407 | #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT) |
408 | |
409 | static inline void set_cold_node(struct page *page, bool is_dir) |
410 | { |
411 | struct f2fs_node *rn = F2FS_NODE(page); |
412 | unsigned int flag = le32_to_cpu(rn->footer.flag); |
413 | |
414 | if (is_dir) |
415 | flag &= ~BIT(COLD_BIT_SHIFT); |
416 | else |
417 | flag |= BIT(COLD_BIT_SHIFT); |
418 | rn->footer.flag = cpu_to_le32(flag); |
419 | } |
420 | |
421 | static inline void set_mark(struct page *page, int mark, int type) |
422 | { |
423 | struct f2fs_node *rn = F2FS_NODE(page); |
424 | unsigned int flag = le32_to_cpu(rn->footer.flag); |
425 | if (mark) |
426 | flag |= BIT(type); |
427 | else |
428 | flag &= ~BIT(type); |
429 | rn->footer.flag = cpu_to_le32(flag); |
430 | |
431 | #ifdef CONFIG_F2FS_CHECK_FS |
432 | f2fs_inode_chksum_set(sbi: F2FS_P_SB(page), page); |
433 | #endif |
434 | } |
435 | #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT) |
436 | #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT) |
437 | |