xref: /openbmc/linux/fs/f2fs/node.h (revision e29dc460)
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 
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 
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 	/* 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 
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)->dirty_nat_cnt >= 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 >= DEF_NAT_CACHE_THRESHOLD;
136 }
137 
138 enum mem_type {
139 	FREE_NIDS,	/* indicates the free nid list */
140 	NAT_ENTRIES,	/* indicates the cached nat entry */
141 	DIRTY_DENTS,	/* indicates dirty dentry pages */
142 	INO_ENTRIES,	/* indicates inode entries */
143 	EXTENT_CACHE,	/* indicates extent cache */
144 	INMEM_PAGES,	/* indicates inmemory pages */
145 	BASE_CHECK,	/* check kernel status */
146 };
147 
148 struct nat_entry_set {
149 	struct list_head set_list;	/* link with other nat sets */
150 	struct list_head entry_list;	/* link with dirty nat entries */
151 	nid_t set;			/* set number*/
152 	unsigned int entry_cnt;		/* the # of nat entries in set */
153 };
154 
155 struct free_nid {
156 	struct list_head list;	/* for free node id list */
157 	nid_t nid;		/* node id */
158 	int state;		/* in use or not: FREE_NID or PREALLOC_NID */
159 };
160 
161 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
162 {
163 	struct f2fs_nm_info *nm_i = NM_I(sbi);
164 	struct free_nid *fnid;
165 
166 	spin_lock(&nm_i->nid_list_lock);
167 	if (nm_i->nid_cnt[FREE_NID] <= 0) {
168 		spin_unlock(&nm_i->nid_list_lock);
169 		return;
170 	}
171 	fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
172 	*nid = fnid->nid;
173 	spin_unlock(&nm_i->nid_list_lock);
174 }
175 
176 /*
177  * inline functions
178  */
179 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
180 {
181 	struct f2fs_nm_info *nm_i = NM_I(sbi);
182 
183 #ifdef CONFIG_F2FS_CHECK_FS
184 	if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
185 						nm_i->bitmap_size))
186 		f2fs_bug_on(sbi, 1);
187 #endif
188 	memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
189 }
190 
191 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
192 {
193 	struct f2fs_nm_info *nm_i = NM_I(sbi);
194 	pgoff_t block_off;
195 	pgoff_t block_addr;
196 
197 	/*
198 	 * block_off = segment_off * 512 + off_in_segment
199 	 * OLD = (segment_off * 512) * 2 + off_in_segment
200 	 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
201 	 */
202 	block_off = NAT_BLOCK_OFFSET(start);
203 
204 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
205 		(block_off << 1) -
206 		(block_off & (sbi->blocks_per_seg - 1)));
207 
208 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
209 		block_addr += sbi->blocks_per_seg;
210 
211 	return block_addr;
212 }
213 
214 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
215 						pgoff_t block_addr)
216 {
217 	struct f2fs_nm_info *nm_i = NM_I(sbi);
218 
219 	block_addr -= nm_i->nat_blkaddr;
220 	block_addr ^= 1 << sbi->log_blocks_per_seg;
221 	return block_addr + nm_i->nat_blkaddr;
222 }
223 
224 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
225 {
226 	unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
227 
228 	f2fs_change_bit(block_off, nm_i->nat_bitmap);
229 #ifdef CONFIG_F2FS_CHECK_FS
230 	f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
231 #endif
232 }
233 
234 static inline nid_t ino_of_node(struct page *node_page)
235 {
236 	struct f2fs_node *rn = F2FS_NODE(node_page);
237 	return le32_to_cpu(rn->footer.ino);
238 }
239 
240 static inline nid_t nid_of_node(struct page *node_page)
241 {
242 	struct f2fs_node *rn = F2FS_NODE(node_page);
243 	return le32_to_cpu(rn->footer.nid);
244 }
245 
246 static inline unsigned int ofs_of_node(struct page *node_page)
247 {
248 	struct f2fs_node *rn = F2FS_NODE(node_page);
249 	unsigned flag = le32_to_cpu(rn->footer.flag);
250 	return flag >> OFFSET_BIT_SHIFT;
251 }
252 
253 static inline __u64 cpver_of_node(struct page *node_page)
254 {
255 	struct f2fs_node *rn = F2FS_NODE(node_page);
256 	return le64_to_cpu(rn->footer.cp_ver);
257 }
258 
259 static inline block_t next_blkaddr_of_node(struct page *node_page)
260 {
261 	struct f2fs_node *rn = F2FS_NODE(node_page);
262 	return le32_to_cpu(rn->footer.next_blkaddr);
263 }
264 
265 static inline void fill_node_footer(struct page *page, nid_t nid,
266 				nid_t ino, unsigned int ofs, bool reset)
267 {
268 	struct f2fs_node *rn = F2FS_NODE(page);
269 	unsigned int old_flag = 0;
270 
271 	if (reset)
272 		memset(rn, 0, sizeof(*rn));
273 	else
274 		old_flag = le32_to_cpu(rn->footer.flag);
275 
276 	rn->footer.nid = cpu_to_le32(nid);
277 	rn->footer.ino = cpu_to_le32(ino);
278 
279 	/* should remain old flag bits such as COLD_BIT_SHIFT */
280 	rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
281 					(old_flag & OFFSET_BIT_MASK));
282 }
283 
284 static inline void copy_node_footer(struct page *dst, struct page *src)
285 {
286 	struct f2fs_node *src_rn = F2FS_NODE(src);
287 	struct f2fs_node *dst_rn = F2FS_NODE(dst);
288 	memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
289 }
290 
291 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
292 {
293 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
294 	struct f2fs_node *rn = F2FS_NODE(page);
295 	__u64 cp_ver = cur_cp_version(ckpt);
296 
297 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
298 		cp_ver |= (cur_cp_crc(ckpt) << 32);
299 
300 	rn->footer.cp_ver = cpu_to_le64(cp_ver);
301 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
302 }
303 
304 static inline bool is_recoverable_dnode(struct page *page)
305 {
306 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
307 	__u64 cp_ver = cur_cp_version(ckpt);
308 
309 	/* Don't care crc part, if fsck.f2fs sets it. */
310 	if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
311 		return (cp_ver << 32) == (cpver_of_node(page) << 32);
312 
313 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
314 		cp_ver |= (cur_cp_crc(ckpt) << 32);
315 
316 	return cp_ver == cpver_of_node(page);
317 }
318 
319 /*
320  * f2fs assigns the following node offsets described as (num).
321  * N = NIDS_PER_BLOCK
322  *
323  *  Inode block (0)
324  *    |- direct node (1)
325  *    |- direct node (2)
326  *    |- indirect node (3)
327  *    |            `- direct node (4 => 4 + N - 1)
328  *    |- indirect node (4 + N)
329  *    |            `- direct node (5 + N => 5 + 2N - 1)
330  *    `- double indirect node (5 + 2N)
331  *                 `- indirect node (6 + 2N)
332  *                       `- direct node
333  *                 ......
334  *                 `- indirect node ((6 + 2N) + x(N + 1))
335  *                       `- direct node
336  *                 ......
337  *                 `- indirect node ((6 + 2N) + (N - 1)(N + 1))
338  *                       `- direct node
339  */
340 static inline bool IS_DNODE(struct page *node_page)
341 {
342 	unsigned int ofs = ofs_of_node(node_page);
343 
344 	if (f2fs_has_xattr_block(ofs))
345 		return true;
346 
347 	if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
348 			ofs == 5 + 2 * NIDS_PER_BLOCK)
349 		return false;
350 	if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
351 		ofs -= 6 + 2 * NIDS_PER_BLOCK;
352 		if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
353 			return false;
354 	}
355 	return true;
356 }
357 
358 static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
359 {
360 	struct f2fs_node *rn = F2FS_NODE(p);
361 
362 	f2fs_wait_on_page_writeback(p, NODE, true);
363 
364 	if (i)
365 		rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
366 	else
367 		rn->in.nid[off] = cpu_to_le32(nid);
368 	return set_page_dirty(p);
369 }
370 
371 static inline nid_t get_nid(struct page *p, int off, bool i)
372 {
373 	struct f2fs_node *rn = F2FS_NODE(p);
374 
375 	if (i)
376 		return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
377 	return le32_to_cpu(rn->in.nid[off]);
378 }
379 
380 /*
381  * Coldness identification:
382  *  - Mark cold files in f2fs_inode_info
383  *  - Mark cold node blocks in their node footer
384  *  - Mark cold data pages in page cache
385  */
386 static inline int is_cold_data(struct page *page)
387 {
388 	return PageChecked(page);
389 }
390 
391 static inline void set_cold_data(struct page *page)
392 {
393 	SetPageChecked(page);
394 }
395 
396 static inline void clear_cold_data(struct page *page)
397 {
398 	ClearPageChecked(page);
399 }
400 
401 static inline int is_node(struct page *page, int type)
402 {
403 	struct f2fs_node *rn = F2FS_NODE(page);
404 	return le32_to_cpu(rn->footer.flag) & (1 << type);
405 }
406 
407 #define is_cold_node(page)	is_node(page, COLD_BIT_SHIFT)
408 #define is_fsync_dnode(page)	is_node(page, FSYNC_BIT_SHIFT)
409 #define is_dent_dnode(page)	is_node(page, DENT_BIT_SHIFT)
410 
411 static inline int is_inline_node(struct page *page)
412 {
413 	return PageChecked(page);
414 }
415 
416 static inline void set_inline_node(struct page *page)
417 {
418 	SetPageChecked(page);
419 }
420 
421 static inline void clear_inline_node(struct page *page)
422 {
423 	ClearPageChecked(page);
424 }
425 
426 static inline void set_cold_node(struct page *page, bool is_dir)
427 {
428 	struct f2fs_node *rn = F2FS_NODE(page);
429 	unsigned int flag = le32_to_cpu(rn->footer.flag);
430 
431 	if (is_dir)
432 		flag &= ~(0x1 << COLD_BIT_SHIFT);
433 	else
434 		flag |= (0x1 << COLD_BIT_SHIFT);
435 	rn->footer.flag = cpu_to_le32(flag);
436 }
437 
438 static inline void set_mark(struct page *page, int mark, int type)
439 {
440 	struct f2fs_node *rn = F2FS_NODE(page);
441 	unsigned int flag = le32_to_cpu(rn->footer.flag);
442 	if (mark)
443 		flag |= (0x1 << type);
444 	else
445 		flag &= ~(0x1 << type);
446 	rn->footer.flag = cpu_to_le32(flag);
447 }
448 #define set_dentry_mark(page, mark)	set_mark(page, mark, DENT_BIT_SHIFT)
449 #define set_fsync_mark(page, mark)	set_mark(page, mark, FSYNC_BIT_SHIFT)
450