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 /* start node id of a node block dedicated to the given node id */
12 #define START_NID(nid) (((nid) / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
13
14 /* node block offset on the NAT area dedicated to the given start node id */
15 #define NAT_BLOCK_OFFSET(start_nid) ((start_nid) / NAT_ENTRY_PER_BLOCK)
16
17 /* # of pages to perform synchronous readahead before building free nids */
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 /* # 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 /* vector size for gang look-up from nat cache that consists of radix tree */
35 #define NATVEC_SIZE 64
36 #define SETVEC_SIZE 32
37
38 /* return value for read_node_page */
39 #define LOCKED_PAGE 1
40
41 /* For flag in struct node_info */
42 enum {
43 IS_CHECKPOINTED, /* is it checkpointed before? */
44 HAS_FSYNCED_INODE, /* is the inode fsynced before? */
45 HAS_LAST_FSYNC, /* has the latest node fsync mark? */
46 IS_DIRTY, /* this nat entry is dirty? */
47 IS_PREALLOC, /* nat entry is preallocated */
48 };
49
50 /*
51 * For node information
52 */
53 struct node_info {
54 nid_t nid; /* node id */
55 nid_t ino; /* inode number of the node's owner */
56 block_t blk_addr; /* block address of the node */
57 unsigned char version; /* version of the node */
58 unsigned char flag; /* for node information bits */
59 };
60
61 struct nat_entry {
62 struct list_head list; /* for clean or dirty nat list */
63 struct node_info ni; /* in-memory node information */
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
copy_node_info(struct node_info * dst,struct node_info * src)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 /* should not copy flag here */
85 }
86
set_nat_flag(struct nat_entry * ne,unsigned int type,bool set)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
get_nat_flag(struct nat_entry * ne,unsigned int type)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
nat_reset_flag(struct nat_entry * ne)103 static inline void nat_reset_flag(struct nat_entry *ne)
104 {
105 /* these states can be set only after checkpoint was done */
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
node_info_from_raw_nat(struct node_info * ni,struct f2fs_nat_entry * raw_ne)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
raw_nat_from_node_info(struct f2fs_nat_entry * raw_ne,struct node_info * ni)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
excess_dirty_nats(struct f2fs_sb_info * sbi)127 static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
128 {
129 return NM_I(sbi)->dirty_nat_cnt >= NM_I(sbi)->max_nid *
130 NM_I(sbi)->dirty_nats_ratio / 100;
131 }
132
excess_cached_nats(struct f2fs_sb_info * sbi)133 static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
134 {
135 return NM_I(sbi)->nat_cnt >= DEF_NAT_CACHE_THRESHOLD;
136 }
137
excess_dirty_nodes(struct f2fs_sb_info * sbi)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, /* indicates the free nid list */
145 NAT_ENTRIES, /* indicates the cached nat entry */
146 DIRTY_DENTS, /* indicates dirty dentry pages */
147 INO_ENTRIES, /* indicates inode entries */
148 EXTENT_CACHE, /* indicates extent cache */
149 INMEM_PAGES, /* indicates inmemory 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
next_free_nid(struct f2fs_sb_info * sbi,nid_t * nid)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 */
get_nat_bitmap(struct f2fs_sb_info * sbi,void * addr)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
current_nat_addr(struct f2fs_sb_info * sbi,nid_t start)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 & (sbi->blocks_per_seg - 1)));
212
213 if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
214 block_addr += sbi->blocks_per_seg;
215
216 return block_addr;
217 }
218
next_nat_addr(struct f2fs_sb_info * sbi,pgoff_t block_addr)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 ^= 1 << sbi->log_blocks_per_seg;
226 return block_addr + nm_i->nat_blkaddr;
227 }
228
set_to_next_nat(struct f2fs_nm_info * nm_i,nid_t start_nid)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
ino_of_node(struct page * node_page)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
nid_of_node(struct page * node_page)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
ofs_of_node(struct page * node_page)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
cpver_of_node(struct page * node_page)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
next_blkaddr_of_node(struct page * node_page)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
fill_node_footer(struct page * page,nid_t nid,nid_t ino,unsigned int ofs,bool reset)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
copy_node_footer(struct page * dst,struct page * src)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
fill_node_footer_blkaddr(struct page * page,block_t blkaddr)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
is_recoverable_dnode(struct page * page)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 */
IS_DNODE(struct page * node_page)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
set_nid(struct page * p,int off,nid_t nid,bool i)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);
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
get_nid(struct page * p,int off,bool i)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 */
is_cold_data(struct page * page)391 static inline int is_cold_data(struct page *page)
392 {
393 return PageChecked(page);
394 }
395
set_cold_data(struct page * page)396 static inline void set_cold_data(struct page *page)
397 {
398 SetPageChecked(page);
399 }
400
clear_cold_data(struct page * page)401 static inline void clear_cold_data(struct page *page)
402 {
403 ClearPageChecked(page);
404 }
405
is_node(struct page * page,int type)406 static inline int is_node(struct page *page, int type)
407 {
408 struct f2fs_node *rn = F2FS_NODE(page);
409 return le32_to_cpu(rn->footer.flag) & (1 << type);
410 }
411
412 #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
413 #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
414 #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
415
is_inline_node(struct page * page)416 static inline int is_inline_node(struct page *page)
417 {
418 return PageChecked(page);
419 }
420
set_inline_node(struct page * page)421 static inline void set_inline_node(struct page *page)
422 {
423 SetPageChecked(page);
424 }
425
clear_inline_node(struct page * page)426 static inline void clear_inline_node(struct page *page)
427 {
428 ClearPageChecked(page);
429 }
430
set_cold_node(struct page * page,bool is_dir)431 static inline void set_cold_node(struct page *page, bool is_dir)
432 {
433 struct f2fs_node *rn = F2FS_NODE(page);
434 unsigned int flag = le32_to_cpu(rn->footer.flag);
435
436 if (is_dir)
437 flag &= ~(0x1 << COLD_BIT_SHIFT);
438 else
439 flag |= (0x1 << COLD_BIT_SHIFT);
440 rn->footer.flag = cpu_to_le32(flag);
441 }
442
set_mark(struct page * page,int mark,int type)443 static inline void set_mark(struct page *page, int mark, int type)
444 {
445 struct f2fs_node *rn = F2FS_NODE(page);
446 unsigned int flag = le32_to_cpu(rn->footer.flag);
447 if (mark)
448 flag |= (0x1 << type);
449 else
450 flag &= ~(0x1 << type);
451 rn->footer.flag = cpu_to_le32(flag);
452
453 #ifdef CONFIG_F2FS_CHECK_FS
454 f2fs_inode_chksum_set(F2FS_P_SB(page), page);
455 #endif
456 }
457 #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
458 #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)
459