xref: /linux/fs/f2fs/node.h (revision 93df8a1ed6231727c5db94a80b1a6bd5ee67cec3)
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 readahead before building free nids */
18 #define FREE_NID_PAGES 4
19 
20 /* maximum readahead size for node during getting data blocks */
21 #define MAX_RA_NODE		128
22 
23 /* control the memory footprint threshold (10MB per 1GB ram) */
24 #define DEF_RAM_THRESHOLD	10
25 
26 /* vector size for gang look-up from nat cache that consists of radix tree */
27 #define NATVEC_SIZE	64
28 #define SETVEC_SIZE	32
29 
30 /* return value for read_node_page */
31 #define LOCKED_PAGE	1
32 
33 /* For flag in struct node_info */
34 enum {
35 	IS_CHECKPOINTED,	/* is it checkpointed before? */
36 	HAS_FSYNCED_INODE,	/* is the inode fsynced before? */
37 	HAS_LAST_FSYNC,		/* has the latest node fsync mark? */
38 	IS_DIRTY,		/* this nat entry is dirty? */
39 };
40 
41 /*
42  * For node information
43  */
44 struct node_info {
45 	nid_t nid;		/* node id */
46 	nid_t ino;		/* inode number of the node's owner */
47 	block_t	blk_addr;	/* block address of the node */
48 	unsigned char version;	/* version of the node */
49 	unsigned char flag;	/* for node information bits */
50 };
51 
52 struct nat_entry {
53 	struct list_head list;	/* for clean or dirty nat list */
54 	struct node_info ni;	/* in-memory node information */
55 };
56 
57 #define nat_get_nid(nat)		(nat->ni.nid)
58 #define nat_set_nid(nat, n)		(nat->ni.nid = n)
59 #define nat_get_blkaddr(nat)		(nat->ni.blk_addr)
60 #define nat_set_blkaddr(nat, b)		(nat->ni.blk_addr = b)
61 #define nat_get_ino(nat)		(nat->ni.ino)
62 #define nat_set_ino(nat, i)		(nat->ni.ino = i)
63 #define nat_get_version(nat)		(nat->ni.version)
64 #define nat_set_version(nat, v)		(nat->ni.version = v)
65 
66 #define inc_node_version(version)	(++version)
67 
68 static inline void copy_node_info(struct node_info *dst,
69 						struct node_info *src)
70 {
71 	dst->nid = src->nid;
72 	dst->ino = src->ino;
73 	dst->blk_addr = src->blk_addr;
74 	dst->version = src->version;
75 	/* should not copy flag here */
76 }
77 
78 static inline void set_nat_flag(struct nat_entry *ne,
79 				unsigned int type, bool set)
80 {
81 	unsigned char mask = 0x01 << type;
82 	if (set)
83 		ne->ni.flag |= mask;
84 	else
85 		ne->ni.flag &= ~mask;
86 }
87 
88 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
89 {
90 	unsigned char mask = 0x01 << type;
91 	return ne->ni.flag & mask;
92 }
93 
94 static inline void nat_reset_flag(struct nat_entry *ne)
95 {
96 	/* these states can be set only after checkpoint was done */
97 	set_nat_flag(ne, IS_CHECKPOINTED, true);
98 	set_nat_flag(ne, HAS_FSYNCED_INODE, false);
99 	set_nat_flag(ne, HAS_LAST_FSYNC, true);
100 }
101 
102 static inline void node_info_from_raw_nat(struct node_info *ni,
103 						struct f2fs_nat_entry *raw_ne)
104 {
105 	ni->ino = le32_to_cpu(raw_ne->ino);
106 	ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
107 	ni->version = raw_ne->version;
108 }
109 
110 static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
111 						struct node_info *ni)
112 {
113 	raw_ne->ino = cpu_to_le32(ni->ino);
114 	raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
115 	raw_ne->version = ni->version;
116 }
117 
118 enum mem_type {
119 	FREE_NIDS,	/* indicates the free nid list */
120 	NAT_ENTRIES,	/* indicates the cached nat entry */
121 	DIRTY_DENTS,	/* indicates dirty dentry pages */
122 	INO_ENTRIES,	/* indicates inode entries */
123 	EXTENT_CACHE,	/* indicates extent cache */
124 	BASE_CHECK,	/* check kernel status */
125 };
126 
127 struct nat_entry_set {
128 	struct list_head set_list;	/* link with other nat sets */
129 	struct list_head entry_list;	/* link with dirty nat entries */
130 	nid_t set;			/* set number*/
131 	unsigned int entry_cnt;		/* the # of nat entries in set */
132 };
133 
134 /*
135  * For free nid mangement
136  */
137 enum nid_state {
138 	NID_NEW,	/* newly added to free nid list */
139 	NID_ALLOC	/* it is allocated */
140 };
141 
142 struct free_nid {
143 	struct list_head list;	/* for free node id list */
144 	nid_t nid;		/* node id */
145 	int state;		/* in use or not: NID_NEW or NID_ALLOC */
146 };
147 
148 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
149 {
150 	struct f2fs_nm_info *nm_i = NM_I(sbi);
151 	struct free_nid *fnid;
152 
153 	spin_lock(&nm_i->free_nid_list_lock);
154 	if (nm_i->fcnt <= 0) {
155 		spin_unlock(&nm_i->free_nid_list_lock);
156 		return;
157 	}
158 	fnid = list_entry(nm_i->free_nid_list.next, struct free_nid, list);
159 	*nid = fnid->nid;
160 	spin_unlock(&nm_i->free_nid_list_lock);
161 }
162 
163 /*
164  * inline functions
165  */
166 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
167 {
168 	struct f2fs_nm_info *nm_i = NM_I(sbi);
169 	memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
170 }
171 
172 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
173 {
174 	struct f2fs_nm_info *nm_i = NM_I(sbi);
175 	pgoff_t block_off;
176 	pgoff_t block_addr;
177 	int seg_off;
178 
179 	block_off = NAT_BLOCK_OFFSET(start);
180 	seg_off = block_off >> sbi->log_blocks_per_seg;
181 
182 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
183 		(seg_off << sbi->log_blocks_per_seg << 1) +
184 		(block_off & ((1 << sbi->log_blocks_per_seg) - 1)));
185 
186 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
187 		block_addr += sbi->blocks_per_seg;
188 
189 	return block_addr;
190 }
191 
192 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
193 						pgoff_t block_addr)
194 {
195 	struct f2fs_nm_info *nm_i = NM_I(sbi);
196 
197 	block_addr -= nm_i->nat_blkaddr;
198 	if ((block_addr >> sbi->log_blocks_per_seg) % 2)
199 		block_addr -= sbi->blocks_per_seg;
200 	else
201 		block_addr += sbi->blocks_per_seg;
202 
203 	return block_addr + nm_i->nat_blkaddr;
204 }
205 
206 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
207 {
208 	unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
209 
210 	f2fs_change_bit(block_off, nm_i->nat_bitmap);
211 }
212 
213 static inline void fill_node_footer(struct page *page, nid_t nid,
214 				nid_t ino, unsigned int ofs, bool reset)
215 {
216 	struct f2fs_node *rn = F2FS_NODE(page);
217 	unsigned int old_flag = 0;
218 
219 	if (reset)
220 		memset(rn, 0, sizeof(*rn));
221 	else
222 		old_flag = le32_to_cpu(rn->footer.flag);
223 
224 	rn->footer.nid = cpu_to_le32(nid);
225 	rn->footer.ino = cpu_to_le32(ino);
226 
227 	/* should remain old flag bits such as COLD_BIT_SHIFT */
228 	rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
229 					(old_flag & OFFSET_BIT_MASK));
230 }
231 
232 static inline void copy_node_footer(struct page *dst, struct page *src)
233 {
234 	struct f2fs_node *src_rn = F2FS_NODE(src);
235 	struct f2fs_node *dst_rn = F2FS_NODE(dst);
236 	memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
237 }
238 
239 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
240 {
241 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
242 	struct f2fs_node *rn = F2FS_NODE(page);
243 
244 	rn->footer.cp_ver = ckpt->checkpoint_ver;
245 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
246 }
247 
248 static inline nid_t ino_of_node(struct page *node_page)
249 {
250 	struct f2fs_node *rn = F2FS_NODE(node_page);
251 	return le32_to_cpu(rn->footer.ino);
252 }
253 
254 static inline nid_t nid_of_node(struct page *node_page)
255 {
256 	struct f2fs_node *rn = F2FS_NODE(node_page);
257 	return le32_to_cpu(rn->footer.nid);
258 }
259 
260 static inline unsigned int ofs_of_node(struct page *node_page)
261 {
262 	struct f2fs_node *rn = F2FS_NODE(node_page);
263 	unsigned flag = le32_to_cpu(rn->footer.flag);
264 	return flag >> OFFSET_BIT_SHIFT;
265 }
266 
267 static inline unsigned long long cpver_of_node(struct page *node_page)
268 {
269 	struct f2fs_node *rn = F2FS_NODE(node_page);
270 	return le64_to_cpu(rn->footer.cp_ver);
271 }
272 
273 static inline block_t next_blkaddr_of_node(struct page *node_page)
274 {
275 	struct f2fs_node *rn = F2FS_NODE(node_page);
276 	return le32_to_cpu(rn->footer.next_blkaddr);
277 }
278 
279 /*
280  * f2fs assigns the following node offsets described as (num).
281  * N = NIDS_PER_BLOCK
282  *
283  *  Inode block (0)
284  *    |- direct node (1)
285  *    |- direct node (2)
286  *    |- indirect node (3)
287  *    |            `- direct node (4 => 4 + N - 1)
288  *    |- indirect node (4 + N)
289  *    |            `- direct node (5 + N => 5 + 2N - 1)
290  *    `- double indirect node (5 + 2N)
291  *                 `- indirect node (6 + 2N)
292  *                       `- direct node
293  *                 ......
294  *                 `- indirect node ((6 + 2N) + x(N + 1))
295  *                       `- direct node
296  *                 ......
297  *                 `- indirect node ((6 + 2N) + (N - 1)(N + 1))
298  *                       `- direct node
299  */
300 static inline bool IS_DNODE(struct page *node_page)
301 {
302 	unsigned int ofs = ofs_of_node(node_page);
303 
304 	if (f2fs_has_xattr_block(ofs))
305 		return false;
306 
307 	if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
308 			ofs == 5 + 2 * NIDS_PER_BLOCK)
309 		return false;
310 	if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
311 		ofs -= 6 + 2 * NIDS_PER_BLOCK;
312 		if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
313 			return false;
314 	}
315 	return true;
316 }
317 
318 static inline void set_nid(struct page *p, int off, nid_t nid, bool i)
319 {
320 	struct f2fs_node *rn = F2FS_NODE(p);
321 
322 	f2fs_wait_on_page_writeback(p, NODE);
323 
324 	if (i)
325 		rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
326 	else
327 		rn->in.nid[off] = cpu_to_le32(nid);
328 	set_page_dirty(p);
329 }
330 
331 static inline nid_t get_nid(struct page *p, int off, bool i)
332 {
333 	struct f2fs_node *rn = F2FS_NODE(p);
334 
335 	if (i)
336 		return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
337 	return le32_to_cpu(rn->in.nid[off]);
338 }
339 
340 /*
341  * Coldness identification:
342  *  - Mark cold files in f2fs_inode_info
343  *  - Mark cold node blocks in their node footer
344  *  - Mark cold data pages in page cache
345  */
346 static inline int is_cold_data(struct page *page)
347 {
348 	return PageChecked(page);
349 }
350 
351 static inline void set_cold_data(struct page *page)
352 {
353 	SetPageChecked(page);
354 }
355 
356 static inline void clear_cold_data(struct page *page)
357 {
358 	ClearPageChecked(page);
359 }
360 
361 static inline int is_node(struct page *page, int type)
362 {
363 	struct f2fs_node *rn = F2FS_NODE(page);
364 	return le32_to_cpu(rn->footer.flag) & (1 << type);
365 }
366 
367 #define is_cold_node(page)	is_node(page, COLD_BIT_SHIFT)
368 #define is_fsync_dnode(page)	is_node(page, FSYNC_BIT_SHIFT)
369 #define is_dent_dnode(page)	is_node(page, DENT_BIT_SHIFT)
370 
371 static inline void set_cold_node(struct inode *inode, struct page *page)
372 {
373 	struct f2fs_node *rn = F2FS_NODE(page);
374 	unsigned int flag = le32_to_cpu(rn->footer.flag);
375 
376 	if (S_ISDIR(inode->i_mode))
377 		flag &= ~(0x1 << COLD_BIT_SHIFT);
378 	else
379 		flag |= (0x1 << COLD_BIT_SHIFT);
380 	rn->footer.flag = cpu_to_le32(flag);
381 }
382 
383 static inline void set_mark(struct page *page, int mark, int type)
384 {
385 	struct f2fs_node *rn = F2FS_NODE(page);
386 	unsigned int flag = le32_to_cpu(rn->footer.flag);
387 	if (mark)
388 		flag |= (0x1 << type);
389 	else
390 		flag &= ~(0x1 << type);
391 	rn->footer.flag = cpu_to_le32(flag);
392 }
393 #define set_dentry_mark(page, mark)	set_mark(page, mark, DENT_BIT_SHIFT)
394 #define set_fsync_mark(page, mark)	set_mark(page, mark, FSYNC_BIT_SHIFT)
395