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