xref: /openbmc/linux/fs/f2fs/node.h (revision dc023681)
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 /* 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 /* check pinned file's alignment status of physical blocks */
42 #define FILE_NOT_ALIGNED	1
43 
44 /* For flag in struct node_info */
45 enum {
46 	IS_CHECKPOINTED,	/* is it checkpointed before? */
47 	HAS_FSYNCED_INODE,	/* is the inode fsynced before? */
48 	HAS_LAST_FSYNC,		/* has the latest node fsync mark? */
49 	IS_DIRTY,		/* this nat entry is dirty? */
50 	IS_PREALLOC,		/* nat entry is preallocated */
51 };
52 
53 /*
54  * For node information
55  */
56 struct node_info {
57 	nid_t nid;		/* node id */
58 	nid_t ino;		/* inode number of the node's owner */
59 	block_t	blk_addr;	/* block address of the node */
60 	unsigned char version;	/* version of the node */
61 	unsigned char flag;	/* for node information bits */
62 };
63 
64 struct nat_entry {
65 	struct list_head list;	/* for clean or dirty nat list */
66 	struct node_info ni;	/* in-memory node information */
67 };
68 
69 #define nat_get_nid(nat)		((nat)->ni.nid)
70 #define nat_set_nid(nat, n)		((nat)->ni.nid = (n))
71 #define nat_get_blkaddr(nat)		((nat)->ni.blk_addr)
72 #define nat_set_blkaddr(nat, b)		((nat)->ni.blk_addr = (b))
73 #define nat_get_ino(nat)		((nat)->ni.ino)
74 #define nat_set_ino(nat, i)		((nat)->ni.ino = (i))
75 #define nat_get_version(nat)		((nat)->ni.version)
76 #define nat_set_version(nat, v)		((nat)->ni.version = (v))
77 
78 #define inc_node_version(version)	(++(version))
79 
80 static inline void copy_node_info(struct node_info *dst,
81 						struct node_info *src)
82 {
83 	dst->nid = src->nid;
84 	dst->ino = src->ino;
85 	dst->blk_addr = src->blk_addr;
86 	dst->version = src->version;
87 	/* should not copy flag here */
88 }
89 
90 static inline void set_nat_flag(struct nat_entry *ne,
91 				unsigned int type, bool set)
92 {
93 	unsigned char mask = 0x01 << type;
94 	if (set)
95 		ne->ni.flag |= mask;
96 	else
97 		ne->ni.flag &= ~mask;
98 }
99 
100 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
101 {
102 	unsigned char mask = 0x01 << type;
103 	return ne->ni.flag & mask;
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 static inline bool excess_dirty_nodes(struct f2fs_sb_info *sbi)
142 {
143 	return get_pages(sbi, F2FS_DIRTY_NODES) >= sbi->blocks_per_seg * 8;
144 }
145 
146 enum mem_type {
147 	FREE_NIDS,	/* indicates the free nid list */
148 	NAT_ENTRIES,	/* indicates the cached nat entry */
149 	DIRTY_DENTS,	/* indicates dirty dentry pages */
150 	INO_ENTRIES,	/* indicates inode entries */
151 	EXTENT_CACHE,	/* indicates extent cache */
152 	INMEM_PAGES,	/* indicates inmemory pages */
153 	DISCARD_CACHE,	/* indicates memory of cached discard cmds */
154 	COMPRESS_PAGE,	/* indicates memory of cached compressed pages */
155 	BASE_CHECK,	/* check kernel status */
156 };
157 
158 struct nat_entry_set {
159 	struct list_head set_list;	/* link with other nat sets */
160 	struct list_head entry_list;	/* link with dirty nat entries */
161 	nid_t set;			/* set number*/
162 	unsigned int entry_cnt;		/* the # of nat entries in set */
163 };
164 
165 struct free_nid {
166 	struct list_head list;	/* for free node id list */
167 	nid_t nid;		/* node id */
168 	int state;		/* in use or not: FREE_NID or PREALLOC_NID */
169 };
170 
171 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
172 {
173 	struct f2fs_nm_info *nm_i = NM_I(sbi);
174 	struct free_nid *fnid;
175 
176 	spin_lock(&nm_i->nid_list_lock);
177 	if (nm_i->nid_cnt[FREE_NID] <= 0) {
178 		spin_unlock(&nm_i->nid_list_lock);
179 		return;
180 	}
181 	fnid = list_first_entry(&nm_i->free_nid_list, struct free_nid, list);
182 	*nid = fnid->nid;
183 	spin_unlock(&nm_i->nid_list_lock);
184 }
185 
186 /*
187  * inline functions
188  */
189 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
190 {
191 	struct f2fs_nm_info *nm_i = NM_I(sbi);
192 
193 #ifdef CONFIG_F2FS_CHECK_FS
194 	if (memcmp(nm_i->nat_bitmap, nm_i->nat_bitmap_mir,
195 						nm_i->bitmap_size))
196 		f2fs_bug_on(sbi, 1);
197 #endif
198 	memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
199 }
200 
201 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
202 {
203 	struct f2fs_nm_info *nm_i = NM_I(sbi);
204 	pgoff_t block_off;
205 	pgoff_t block_addr;
206 
207 	/*
208 	 * block_off = segment_off * 512 + off_in_segment
209 	 * OLD = (segment_off * 512) * 2 + off_in_segment
210 	 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
211 	 */
212 	block_off = NAT_BLOCK_OFFSET(start);
213 
214 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
215 		(block_off << 1) -
216 		(block_off & (sbi->blocks_per_seg - 1)));
217 
218 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
219 		block_addr += sbi->blocks_per_seg;
220 
221 	return block_addr;
222 }
223 
224 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
225 						pgoff_t block_addr)
226 {
227 	struct f2fs_nm_info *nm_i = NM_I(sbi);
228 
229 	block_addr -= nm_i->nat_blkaddr;
230 	block_addr ^= 1 << sbi->log_blocks_per_seg;
231 	return block_addr + nm_i->nat_blkaddr;
232 }
233 
234 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
235 {
236 	unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
237 
238 	f2fs_change_bit(block_off, nm_i->nat_bitmap);
239 #ifdef CONFIG_F2FS_CHECK_FS
240 	f2fs_change_bit(block_off, nm_i->nat_bitmap_mir);
241 #endif
242 }
243 
244 static inline nid_t ino_of_node(struct page *node_page)
245 {
246 	struct f2fs_node *rn = F2FS_NODE(node_page);
247 	return le32_to_cpu(rn->footer.ino);
248 }
249 
250 static inline nid_t nid_of_node(struct page *node_page)
251 {
252 	struct f2fs_node *rn = F2FS_NODE(node_page);
253 	return le32_to_cpu(rn->footer.nid);
254 }
255 
256 static inline unsigned int ofs_of_node(struct page *node_page)
257 {
258 	struct f2fs_node *rn = F2FS_NODE(node_page);
259 	unsigned flag = le32_to_cpu(rn->footer.flag);
260 	return flag >> OFFSET_BIT_SHIFT;
261 }
262 
263 static inline __u64 cpver_of_node(struct page *node_page)
264 {
265 	struct f2fs_node *rn = F2FS_NODE(node_page);
266 	return le64_to_cpu(rn->footer.cp_ver);
267 }
268 
269 static inline block_t next_blkaddr_of_node(struct page *node_page)
270 {
271 	struct f2fs_node *rn = F2FS_NODE(node_page);
272 	return le32_to_cpu(rn->footer.next_blkaddr);
273 }
274 
275 static inline void fill_node_footer(struct page *page, nid_t nid,
276 				nid_t ino, unsigned int ofs, bool reset)
277 {
278 	struct f2fs_node *rn = F2FS_NODE(page);
279 	unsigned int old_flag = 0;
280 
281 	if (reset)
282 		memset(rn, 0, sizeof(*rn));
283 	else
284 		old_flag = le32_to_cpu(rn->footer.flag);
285 
286 	rn->footer.nid = cpu_to_le32(nid);
287 	rn->footer.ino = cpu_to_le32(ino);
288 
289 	/* should remain old flag bits such as COLD_BIT_SHIFT */
290 	rn->footer.flag = cpu_to_le32((ofs << OFFSET_BIT_SHIFT) |
291 					(old_flag & OFFSET_BIT_MASK));
292 }
293 
294 static inline void copy_node_footer(struct page *dst, struct page *src)
295 {
296 	struct f2fs_node *src_rn = F2FS_NODE(src);
297 	struct f2fs_node *dst_rn = F2FS_NODE(dst);
298 	memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
299 }
300 
301 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
302 {
303 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
304 	struct f2fs_node *rn = F2FS_NODE(page);
305 	__u64 cp_ver = cur_cp_version(ckpt);
306 
307 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
308 		cp_ver |= (cur_cp_crc(ckpt) << 32);
309 
310 	rn->footer.cp_ver = cpu_to_le64(cp_ver);
311 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
312 }
313 
314 static inline bool is_recoverable_dnode(struct page *page)
315 {
316 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
317 	__u64 cp_ver = cur_cp_version(ckpt);
318 
319 	/* Don't care crc part, if fsck.f2fs sets it. */
320 	if (__is_set_ckpt_flags(ckpt, CP_NOCRC_RECOVERY_FLAG))
321 		return (cp_ver << 32) == (cpver_of_node(page) << 32);
322 
323 	if (__is_set_ckpt_flags(ckpt, CP_CRC_RECOVERY_FLAG))
324 		cp_ver |= (cur_cp_crc(ckpt) << 32);
325 
326 	return cp_ver == cpver_of_node(page);
327 }
328 
329 /*
330  * f2fs assigns the following node offsets described as (num).
331  * N = NIDS_PER_BLOCK
332  *
333  *  Inode block (0)
334  *    |- direct node (1)
335  *    |- direct node (2)
336  *    |- indirect node (3)
337  *    |            `- direct node (4 => 4 + N - 1)
338  *    |- indirect node (4 + N)
339  *    |            `- direct node (5 + N => 5 + 2N - 1)
340  *    `- double indirect node (5 + 2N)
341  *                 `- indirect node (6 + 2N)
342  *                       `- direct node
343  *                 ......
344  *                 `- indirect node ((6 + 2N) + x(N + 1))
345  *                       `- direct node
346  *                 ......
347  *                 `- indirect node ((6 + 2N) + (N - 1)(N + 1))
348  *                       `- direct node
349  */
350 static inline bool IS_DNODE(struct page *node_page)
351 {
352 	unsigned int ofs = ofs_of_node(node_page);
353 
354 	if (f2fs_has_xattr_block(ofs))
355 		return true;
356 
357 	if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
358 			ofs == 5 + 2 * NIDS_PER_BLOCK)
359 		return false;
360 	if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
361 		ofs -= 6 + 2 * NIDS_PER_BLOCK;
362 		if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
363 			return false;
364 	}
365 	return true;
366 }
367 
368 static inline int set_nid(struct page *p, int off, nid_t nid, bool i)
369 {
370 	struct f2fs_node *rn = F2FS_NODE(p);
371 
372 	f2fs_wait_on_page_writeback(p, NODE, true, true);
373 
374 	if (i)
375 		rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
376 	else
377 		rn->in.nid[off] = cpu_to_le32(nid);
378 	return set_page_dirty(p);
379 }
380 
381 static inline nid_t get_nid(struct page *p, int off, bool i)
382 {
383 	struct f2fs_node *rn = F2FS_NODE(p);
384 
385 	if (i)
386 		return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
387 	return le32_to_cpu(rn->in.nid[off]);
388 }
389 
390 /*
391  * Coldness identification:
392  *  - Mark cold files in f2fs_inode_info
393  *  - Mark cold node blocks in their node footer
394  *  - Mark cold data pages in page cache
395  */
396 
397 static inline int is_node(struct page *page, int type)
398 {
399 	struct f2fs_node *rn = F2FS_NODE(page);
400 	return le32_to_cpu(rn->footer.flag) & (1 << type);
401 }
402 
403 #define is_cold_node(page)	is_node(page, COLD_BIT_SHIFT)
404 #define is_fsync_dnode(page)	is_node(page, FSYNC_BIT_SHIFT)
405 #define is_dent_dnode(page)	is_node(page, DENT_BIT_SHIFT)
406 
407 static inline void set_cold_node(struct page *page, bool is_dir)
408 {
409 	struct f2fs_node *rn = F2FS_NODE(page);
410 	unsigned int flag = le32_to_cpu(rn->footer.flag);
411 
412 	if (is_dir)
413 		flag &= ~(0x1 << COLD_BIT_SHIFT);
414 	else
415 		flag |= (0x1 << COLD_BIT_SHIFT);
416 	rn->footer.flag = cpu_to_le32(flag);
417 }
418 
419 static inline void set_mark(struct page *page, int mark, int type)
420 {
421 	struct f2fs_node *rn = F2FS_NODE(page);
422 	unsigned int flag = le32_to_cpu(rn->footer.flag);
423 	if (mark)
424 		flag |= (0x1 << type);
425 	else
426 		flag &= ~(0x1 << type);
427 	rn->footer.flag = cpu_to_le32(flag);
428 
429 #ifdef CONFIG_F2FS_CHECK_FS
430 	f2fs_inode_chksum_set(F2FS_P_SB(page), page);
431 #endif
432 }
433 #define set_dentry_mark(page, mark)	set_mark(page, mark, DENT_BIT_SHIFT)
434 #define set_fsync_mark(page, mark)	set_mark(page, mark, FSYNC_BIT_SHIFT)
435