4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
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)
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)
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)
21 #define DEF_RA_NID_PAGES 0 /* # of nid pages to be readaheaded */
23 /* maximum readahead size for node during getting data blocks */
24 #define MAX_RA_NODE 128
26 /* control the memory footprint threshold (10MB per 1GB ram) */
27 #define DEF_RAM_THRESHOLD 1
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
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
38 /* return value for read_node_page */
41 /* For flag in struct node_info */
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 */
51 * For node information
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 */
62 struct list_head list
; /* for clean or dirty nat list */
63 struct node_info ni
; /* in-memory node information */
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))
75 #define inc_node_version(version) (++(version))
77 static inline void copy_node_info(struct node_info
*dst
,
78 struct node_info
*src
)
82 dst
->blk_addr
= src
->blk_addr
;
83 dst
->version
= src
->version
;
84 /* should not copy flag here */
87 static inline void set_nat_flag(struct nat_entry
*ne
,
88 unsigned int type
, bool set
)
90 unsigned char mask
= 0x01 << type
;
97 static inline bool get_nat_flag(struct nat_entry
*ne
, unsigned int type
)
99 unsigned char mask
= 0x01 << type
;
100 return ne
->ni
.flag
& mask
;
103 static inline void nat_reset_flag(struct nat_entry
*ne
)
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);
111 static inline void node_info_from_raw_nat(struct node_info
*ni
,
112 struct f2fs_nat_entry
*raw_ne
)
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
;
119 static inline void raw_nat_from_node_info(struct f2fs_nat_entry
*raw_ne
,
120 struct node_info
*ni
)
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
;
127 static inline bool excess_dirty_nats(struct f2fs_sb_info
*sbi
)
129 return NM_I(sbi
)->dirty_nat_cnt
>= NM_I(sbi
)->max_nid
*
130 NM_I(sbi
)->dirty_nats_ratio
/ 100;
133 static inline bool excess_cached_nats(struct f2fs_sb_info
*sbi
)
135 return NM_I(sbi
)->nat_cnt
>= DEF_NAT_CACHE_THRESHOLD
;
138 static inline bool excess_dirty_nodes(struct f2fs_sb_info
*sbi
)
140 return get_pages(sbi
, F2FS_DIRTY_NODES
) >= sbi
->blocks_per_seg
* 8;
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 */
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 */
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 */
166 static inline void next_free_nid(struct f2fs_sb_info
*sbi
, nid_t
*nid
)
168 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
169 struct free_nid
*fnid
;
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
);
176 fnid
= list_first_entry(&nm_i
->free_nid_list
, struct free_nid
, list
);
178 spin_unlock(&nm_i
->nid_list_lock
);
184 static inline void get_nat_bitmap(struct f2fs_sb_info
*sbi
, void *addr
)
186 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
188 #ifdef CONFIG_F2FS_CHECK_FS
189 if (memcmp(nm_i
->nat_bitmap
, nm_i
->nat_bitmap_mir
,
193 memcpy(addr
, nm_i
->nat_bitmap
, nm_i
->bitmap_size
);
196 static inline pgoff_t
current_nat_addr(struct f2fs_sb_info
*sbi
, nid_t start
)
198 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
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
207 block_off
= NAT_BLOCK_OFFSET(start
);
209 block_addr
= (pgoff_t
)(nm_i
->nat_blkaddr
+
211 (block_off
& (sbi
->blocks_per_seg
- 1)));
213 if (f2fs_test_bit(block_off
, nm_i
->nat_bitmap
))
214 block_addr
+= sbi
->blocks_per_seg
;
219 static inline pgoff_t
next_nat_addr(struct f2fs_sb_info
*sbi
,
222 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
224 block_addr
-= nm_i
->nat_blkaddr
;
225 block_addr
^= 1 << sbi
->log_blocks_per_seg
;
226 return block_addr
+ nm_i
->nat_blkaddr
;
229 static inline void set_to_next_nat(struct f2fs_nm_info
*nm_i
, nid_t start_nid
)
231 unsigned int block_off
= NAT_BLOCK_OFFSET(start_nid
);
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
);
239 static inline nid_t
ino_of_node(struct page
*node_page
)
241 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
242 return le32_to_cpu(rn
->footer
.ino
);
245 static inline nid_t
nid_of_node(struct page
*node_page
)
247 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
248 return le32_to_cpu(rn
->footer
.nid
);
251 static inline unsigned int ofs_of_node(struct page
*node_page
)
253 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
254 unsigned flag
= le32_to_cpu(rn
->footer
.flag
);
255 return flag
>> OFFSET_BIT_SHIFT
;
258 static inline __u64
cpver_of_node(struct page
*node_page
)
260 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
261 return le64_to_cpu(rn
->footer
.cp_ver
);
264 static inline block_t
next_blkaddr_of_node(struct page
*node_page
)
266 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
267 return le32_to_cpu(rn
->footer
.next_blkaddr
);
270 static inline void fill_node_footer(struct page
*page
, nid_t nid
,
271 nid_t ino
, unsigned int ofs
, bool reset
)
273 struct f2fs_node
*rn
= F2FS_NODE(page
);
274 unsigned int old_flag
= 0;
277 memset(rn
, 0, sizeof(*rn
));
279 old_flag
= le32_to_cpu(rn
->footer
.flag
);
281 rn
->footer
.nid
= cpu_to_le32(nid
);
282 rn
->footer
.ino
= cpu_to_le32(ino
);
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
));
289 static inline void copy_node_footer(struct page
*dst
, struct page
*src
)
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
));
296 static inline void fill_node_footer_blkaddr(struct page
*page
, block_t blkaddr
)
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
);
302 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
303 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
305 rn
->footer
.cp_ver
= cpu_to_le64(cp_ver
);
306 rn
->footer
.next_blkaddr
= cpu_to_le32(blkaddr
);
309 static inline bool is_recoverable_dnode(struct page
*page
)
311 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
312 __u64 cp_ver
= cur_cp_version(ckpt
);
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);
318 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
319 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
321 return cp_ver
== cpver_of_node(page
);
325 * f2fs assigns the following node offsets described as (num).
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)
339 * `- indirect node ((6 + 2N) + x(N + 1))
342 * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
345 static inline bool IS_DNODE(struct page
*node_page
)
347 unsigned int ofs
= ofs_of_node(node_page
);
349 if (f2fs_has_xattr_block(ofs
))
352 if (ofs
== 3 || ofs
== 4 + NIDS_PER_BLOCK
||
353 ofs
== 5 + 2 * NIDS_PER_BLOCK
)
355 if (ofs
>= 6 + 2 * NIDS_PER_BLOCK
) {
356 ofs
-= 6 + 2 * NIDS_PER_BLOCK
;
357 if (!((long int)ofs
% (NIDS_PER_BLOCK
+ 1)))
363 static inline int set_nid(struct page
*p
, int off
, nid_t nid
, bool i
)
365 struct f2fs_node
*rn
= F2FS_NODE(p
);
367 f2fs_wait_on_page_writeback(p
, NODE
, true);
370 rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
] = cpu_to_le32(nid
);
372 rn
->in
.nid
[off
] = cpu_to_le32(nid
);
373 return set_page_dirty(p
);
376 static inline nid_t
get_nid(struct page
*p
, int off
, bool i
)
378 struct f2fs_node
*rn
= F2FS_NODE(p
);
381 return le32_to_cpu(rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
]);
382 return le32_to_cpu(rn
->in
.nid
[off
]);
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
391 static inline int is_cold_data(struct page
*page
)
393 return PageChecked(page
);
396 static inline void set_cold_data(struct page
*page
)
398 SetPageChecked(page
);
401 static inline void clear_cold_data(struct page
*page
)
403 ClearPageChecked(page
);
406 static inline int is_node(struct page
*page
, int type
)
408 struct f2fs_node
*rn
= F2FS_NODE(page
);
409 return le32_to_cpu(rn
->footer
.flag
) & (1 << type
);
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)
416 static inline int is_inline_node(struct page
*page
)
418 return PageChecked(page
);
421 static inline void set_inline_node(struct page
*page
)
423 SetPageChecked(page
);
426 static inline void clear_inline_node(struct page
*page
)
428 ClearPageChecked(page
);
431 static inline void set_cold_node(struct page
*page
, bool is_dir
)
433 struct f2fs_node
*rn
= F2FS_NODE(page
);
434 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
437 flag
&= ~(0x1 << COLD_BIT_SHIFT
);
439 flag
|= (0x1 << COLD_BIT_SHIFT
);
440 rn
->footer
.flag
= cpu_to_le32(flag
);
443 static inline void set_mark(struct page
*page
, int mark
, int type
)
445 struct f2fs_node
*rn
= F2FS_NODE(page
);
446 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
448 flag
|= (0x1 << type
);
450 flag
&= ~(0x1 << type
);
451 rn
->footer
.flag
= cpu_to_le32(flag
);
453 #ifdef CONFIG_F2FS_CHECK_FS
454 f2fs_inode_chksum_set(F2FS_P_SB(page
), page
);
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)