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? */
50 * For node information
53 nid_t nid
; /* node id */
54 nid_t ino
; /* inode number of the node's owner */
55 block_t blk_addr
; /* block address of the node */
56 unsigned char version
; /* version of the node */
57 unsigned char flag
; /* for node information bits */
61 struct list_head list
; /* for clean or dirty nat list */
62 struct node_info ni
; /* in-memory node information */
65 #define nat_get_nid(nat) (nat->ni.nid)
66 #define nat_set_nid(nat, n) (nat->ni.nid = n)
67 #define nat_get_blkaddr(nat) (nat->ni.blk_addr)
68 #define nat_set_blkaddr(nat, b) (nat->ni.blk_addr = b)
69 #define nat_get_ino(nat) (nat->ni.ino)
70 #define nat_set_ino(nat, i) (nat->ni.ino = i)
71 #define nat_get_version(nat) (nat->ni.version)
72 #define nat_set_version(nat, v) (nat->ni.version = v)
74 #define inc_node_version(version) (++version)
76 static inline void copy_node_info(struct node_info
*dst
,
77 struct node_info
*src
)
81 dst
->blk_addr
= src
->blk_addr
;
82 dst
->version
= src
->version
;
83 /* should not copy flag here */
86 static inline void set_nat_flag(struct nat_entry
*ne
,
87 unsigned int type
, bool set
)
89 unsigned char mask
= 0x01 << type
;
96 static inline bool get_nat_flag(struct nat_entry
*ne
, unsigned int type
)
98 unsigned char mask
= 0x01 << type
;
99 return ne
->ni
.flag
& mask
;
102 static inline void nat_reset_flag(struct nat_entry
*ne
)
104 /* these states can be set only after checkpoint was done */
105 set_nat_flag(ne
, IS_CHECKPOINTED
, true);
106 set_nat_flag(ne
, HAS_FSYNCED_INODE
, false);
107 set_nat_flag(ne
, HAS_LAST_FSYNC
, true);
110 static inline void node_info_from_raw_nat(struct node_info
*ni
,
111 struct f2fs_nat_entry
*raw_ne
)
113 ni
->ino
= le32_to_cpu(raw_ne
->ino
);
114 ni
->blk_addr
= le32_to_cpu(raw_ne
->block_addr
);
115 ni
->version
= raw_ne
->version
;
118 static inline void raw_nat_from_node_info(struct f2fs_nat_entry
*raw_ne
,
119 struct node_info
*ni
)
121 raw_ne
->ino
= cpu_to_le32(ni
->ino
);
122 raw_ne
->block_addr
= cpu_to_le32(ni
->blk_addr
);
123 raw_ne
->version
= ni
->version
;
126 static inline bool excess_dirty_nats(struct f2fs_sb_info
*sbi
)
128 return NM_I(sbi
)->dirty_nat_cnt
>= NM_I(sbi
)->max_nid
*
129 NM_I(sbi
)->dirty_nats_ratio
/ 100;
132 static inline bool excess_cached_nats(struct f2fs_sb_info
*sbi
)
134 return NM_I(sbi
)->nat_cnt
>= DEF_NAT_CACHE_THRESHOLD
;
138 FREE_NIDS
, /* indicates the free nid list */
139 NAT_ENTRIES
, /* indicates the cached nat entry */
140 DIRTY_DENTS
, /* indicates dirty dentry pages */
141 INO_ENTRIES
, /* indicates inode entries */
142 EXTENT_CACHE
, /* indicates extent cache */
143 BASE_CHECK
, /* check kernel status */
146 struct nat_entry_set
{
147 struct list_head set_list
; /* link with other nat sets */
148 struct list_head entry_list
; /* link with dirty nat entries */
149 nid_t set
; /* set number*/
150 unsigned int entry_cnt
; /* the # of nat entries in set */
154 * For free nid mangement
157 NID_NEW
, /* newly added to free nid list */
158 NID_ALLOC
/* it is allocated */
162 struct list_head list
; /* for free node id list */
163 nid_t nid
; /* node id */
164 int state
; /* in use or not: NID_NEW or NID_ALLOC */
167 static inline void next_free_nid(struct f2fs_sb_info
*sbi
, nid_t
*nid
)
169 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
170 struct free_nid
*fnid
;
172 spin_lock(&nm_i
->free_nid_list_lock
);
173 if (nm_i
->fcnt
<= 0) {
174 spin_unlock(&nm_i
->free_nid_list_lock
);
177 fnid
= list_entry(nm_i
->free_nid_list
.next
, struct free_nid
, list
);
179 spin_unlock(&nm_i
->free_nid_list_lock
);
185 static inline void get_nat_bitmap(struct f2fs_sb_info
*sbi
, void *addr
)
187 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
188 memcpy(addr
, nm_i
->nat_bitmap
, nm_i
->bitmap_size
);
191 static inline pgoff_t
current_nat_addr(struct f2fs_sb_info
*sbi
, nid_t start
)
193 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
198 block_off
= NAT_BLOCK_OFFSET(start
);
199 seg_off
= block_off
>> sbi
->log_blocks_per_seg
;
201 block_addr
= (pgoff_t
)(nm_i
->nat_blkaddr
+
202 (seg_off
<< sbi
->log_blocks_per_seg
<< 1) +
203 (block_off
& (sbi
->blocks_per_seg
- 1)));
205 if (f2fs_test_bit(block_off
, nm_i
->nat_bitmap
))
206 block_addr
+= sbi
->blocks_per_seg
;
211 static inline pgoff_t
next_nat_addr(struct f2fs_sb_info
*sbi
,
214 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
216 block_addr
-= nm_i
->nat_blkaddr
;
217 if ((block_addr
>> sbi
->log_blocks_per_seg
) % 2)
218 block_addr
-= sbi
->blocks_per_seg
;
220 block_addr
+= sbi
->blocks_per_seg
;
222 return block_addr
+ nm_i
->nat_blkaddr
;
225 static inline void set_to_next_nat(struct f2fs_nm_info
*nm_i
, nid_t start_nid
)
227 unsigned int block_off
= NAT_BLOCK_OFFSET(start_nid
);
229 f2fs_change_bit(block_off
, nm_i
->nat_bitmap
);
232 static inline nid_t
ino_of_node(struct page
*node_page
)
234 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
235 return le32_to_cpu(rn
->footer
.ino
);
238 static inline nid_t
nid_of_node(struct page
*node_page
)
240 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
241 return le32_to_cpu(rn
->footer
.nid
);
244 static inline unsigned int ofs_of_node(struct page
*node_page
)
246 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
247 unsigned flag
= le32_to_cpu(rn
->footer
.flag
);
248 return flag
>> OFFSET_BIT_SHIFT
;
251 static inline __u64
cpver_of_node(struct page
*node_page
)
253 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
254 return le64_to_cpu(rn
->footer
.cp_ver
);
257 static inline block_t
next_blkaddr_of_node(struct page
*node_page
)
259 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
260 return le32_to_cpu(rn
->footer
.next_blkaddr
);
263 static inline void fill_node_footer(struct page
*page
, nid_t nid
,
264 nid_t ino
, unsigned int ofs
, bool reset
)
266 struct f2fs_node
*rn
= F2FS_NODE(page
);
267 unsigned int old_flag
= 0;
270 memset(rn
, 0, sizeof(*rn
));
272 old_flag
= le32_to_cpu(rn
->footer
.flag
);
274 rn
->footer
.nid
= cpu_to_le32(nid
);
275 rn
->footer
.ino
= cpu_to_le32(ino
);
277 /* should remain old flag bits such as COLD_BIT_SHIFT */
278 rn
->footer
.flag
= cpu_to_le32((ofs
<< OFFSET_BIT_SHIFT
) |
279 (old_flag
& OFFSET_BIT_MASK
));
282 static inline void copy_node_footer(struct page
*dst
, struct page
*src
)
284 struct f2fs_node
*src_rn
= F2FS_NODE(src
);
285 struct f2fs_node
*dst_rn
= F2FS_NODE(dst
);
286 memcpy(&dst_rn
->footer
, &src_rn
->footer
, sizeof(struct node_footer
));
289 static inline void fill_node_footer_blkaddr(struct page
*page
, block_t blkaddr
)
291 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
292 struct f2fs_node
*rn
= F2FS_NODE(page
);
293 size_t crc_offset
= le32_to_cpu(ckpt
->checksum_offset
);
294 __u64 cp_ver
= le64_to_cpu(ckpt
->checkpoint_ver
);
296 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
)) {
297 __u64 crc
= le32_to_cpu(*((__le32
*)
298 ((unsigned char *)ckpt
+ crc_offset
)));
299 cp_ver
|= (crc
<< 32);
301 rn
->footer
.cp_ver
= cpu_to_le64(cp_ver
);
302 rn
->footer
.next_blkaddr
= cpu_to_le32(blkaddr
);
305 static inline bool is_recoverable_dnode(struct page
*page
)
307 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
308 size_t crc_offset
= le32_to_cpu(ckpt
->checksum_offset
);
309 __u64 cp_ver
= cur_cp_version(ckpt
);
311 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
)) {
312 __u64 crc
= le32_to_cpu(*((__le32
*)
313 ((unsigned char *)ckpt
+ crc_offset
)));
314 cp_ver
|= (crc
<< 32);
316 return cpu_to_le64(cp_ver
) == cpver_of_node(page
);
320 * f2fs assigns the following node offsets described as (num).
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)
334 * `- indirect node ((6 + 2N) + x(N + 1))
337 * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
340 static inline bool IS_DNODE(struct page
*node_page
)
342 unsigned int ofs
= ofs_of_node(node_page
);
344 if (f2fs_has_xattr_block(ofs
))
347 if (ofs
== 3 || ofs
== 4 + NIDS_PER_BLOCK
||
348 ofs
== 5 + 2 * NIDS_PER_BLOCK
)
350 if (ofs
>= 6 + 2 * NIDS_PER_BLOCK
) {
351 ofs
-= 6 + 2 * NIDS_PER_BLOCK
;
352 if (!((long int)ofs
% (NIDS_PER_BLOCK
+ 1)))
358 static inline int set_nid(struct page
*p
, int off
, nid_t nid
, bool i
)
360 struct f2fs_node
*rn
= F2FS_NODE(p
);
362 f2fs_wait_on_page_writeback(p
, NODE
, true);
365 rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
] = cpu_to_le32(nid
);
367 rn
->in
.nid
[off
] = cpu_to_le32(nid
);
368 return set_page_dirty(p
);
371 static inline nid_t
get_nid(struct page
*p
, int off
, bool i
)
373 struct f2fs_node
*rn
= F2FS_NODE(p
);
376 return le32_to_cpu(rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
]);
377 return le32_to_cpu(rn
->in
.nid
[off
]);
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
386 static inline int is_cold_data(struct page
*page
)
388 return PageChecked(page
);
391 static inline void set_cold_data(struct page
*page
)
393 SetPageChecked(page
);
396 static inline void clear_cold_data(struct page
*page
)
398 ClearPageChecked(page
);
401 static inline int is_node(struct page
*page
, int type
)
403 struct f2fs_node
*rn
= F2FS_NODE(page
);
404 return le32_to_cpu(rn
->footer
.flag
) & (1 << type
);
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)
411 static inline int is_inline_node(struct page
*page
)
413 return PageChecked(page
);
416 static inline void set_inline_node(struct page
*page
)
418 SetPageChecked(page
);
421 static inline void clear_inline_node(struct page
*page
)
423 ClearPageChecked(page
);
426 static inline void set_cold_node(struct inode
*inode
, struct page
*page
)
428 struct f2fs_node
*rn
= F2FS_NODE(page
);
429 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
431 if (S_ISDIR(inode
->i_mode
))
432 flag
&= ~(0x1 << COLD_BIT_SHIFT
);
434 flag
|= (0x1 << COLD_BIT_SHIFT
);
435 rn
->footer
.flag
= cpu_to_le32(flag
);
438 static inline void set_mark(struct page
*page
, int mark
, int type
)
440 struct f2fs_node
*rn
= F2FS_NODE(page
);
441 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
443 flag
|= (0x1 << type
);
445 flag
&= ~(0x1 << type
);
446 rn
->footer
.flag
= cpu_to_le32(flag
);
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)