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 INMEM_PAGES
, /* indicates inmemory pages */
144 BASE_CHECK
, /* check kernel status */
147 struct nat_entry_set
{
148 struct list_head set_list
; /* link with other nat sets */
149 struct list_head entry_list
; /* link with dirty nat entries */
150 nid_t set
; /* set number*/
151 unsigned int entry_cnt
; /* the # of nat entries in set */
155 struct list_head list
; /* for free node id list */
156 nid_t nid
; /* node id */
157 int state
; /* in use or not: FREE_NID or PREALLOC_NID */
160 static inline void next_free_nid(struct f2fs_sb_info
*sbi
, nid_t
*nid
)
162 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
163 struct free_nid
*fnid
;
165 spin_lock(&nm_i
->nid_list_lock
);
166 if (nm_i
->nid_cnt
[FREE_NID
] <= 0) {
167 spin_unlock(&nm_i
->nid_list_lock
);
170 fnid
= list_first_entry(&nm_i
->free_nid_list
, struct free_nid
, list
);
172 spin_unlock(&nm_i
->nid_list_lock
);
178 static inline void get_nat_bitmap(struct f2fs_sb_info
*sbi
, void *addr
)
180 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
182 #ifdef CONFIG_F2FS_CHECK_FS
183 if (memcmp(nm_i
->nat_bitmap
, nm_i
->nat_bitmap_mir
,
187 memcpy(addr
, nm_i
->nat_bitmap
, nm_i
->bitmap_size
);
190 static inline pgoff_t
current_nat_addr(struct f2fs_sb_info
*sbi
, nid_t start
)
192 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
197 * block_off = segment_off * 512 + off_in_segment
198 * OLD = (segment_off * 512) * 2 + off_in_segment
199 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
201 block_off
= NAT_BLOCK_OFFSET(start
);
203 block_addr
= (pgoff_t
)(nm_i
->nat_blkaddr
+
205 (block_off
& (sbi
->blocks_per_seg
- 1)));
207 if (f2fs_test_bit(block_off
, nm_i
->nat_bitmap
))
208 block_addr
+= sbi
->blocks_per_seg
;
213 static inline pgoff_t
next_nat_addr(struct f2fs_sb_info
*sbi
,
216 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
218 block_addr
-= nm_i
->nat_blkaddr
;
219 block_addr
^= 1 << sbi
->log_blocks_per_seg
;
220 return block_addr
+ nm_i
->nat_blkaddr
;
223 static inline void set_to_next_nat(struct f2fs_nm_info
*nm_i
, nid_t start_nid
)
225 unsigned int block_off
= NAT_BLOCK_OFFSET(start_nid
);
227 f2fs_change_bit(block_off
, nm_i
->nat_bitmap
);
228 #ifdef CONFIG_F2FS_CHECK_FS
229 f2fs_change_bit(block_off
, nm_i
->nat_bitmap_mir
);
233 static inline nid_t
ino_of_node(struct page
*node_page
)
235 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
236 return le32_to_cpu(rn
->footer
.ino
);
239 static inline nid_t
nid_of_node(struct page
*node_page
)
241 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
242 return le32_to_cpu(rn
->footer
.nid
);
245 static inline unsigned int ofs_of_node(struct page
*node_page
)
247 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
248 unsigned flag
= le32_to_cpu(rn
->footer
.flag
);
249 return flag
>> OFFSET_BIT_SHIFT
;
252 static inline __u64
cpver_of_node(struct page
*node_page
)
254 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
255 return le64_to_cpu(rn
->footer
.cp_ver
);
258 static inline block_t
next_blkaddr_of_node(struct page
*node_page
)
260 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
261 return le32_to_cpu(rn
->footer
.next_blkaddr
);
264 static inline void fill_node_footer(struct page
*page
, nid_t nid
,
265 nid_t ino
, unsigned int ofs
, bool reset
)
267 struct f2fs_node
*rn
= F2FS_NODE(page
);
268 unsigned int old_flag
= 0;
271 memset(rn
, 0, sizeof(*rn
));
273 old_flag
= le32_to_cpu(rn
->footer
.flag
);
275 rn
->footer
.nid
= cpu_to_le32(nid
);
276 rn
->footer
.ino
= cpu_to_le32(ino
);
278 /* should remain old flag bits such as COLD_BIT_SHIFT */
279 rn
->footer
.flag
= cpu_to_le32((ofs
<< OFFSET_BIT_SHIFT
) |
280 (old_flag
& OFFSET_BIT_MASK
));
283 static inline void copy_node_footer(struct page
*dst
, struct page
*src
)
285 struct f2fs_node
*src_rn
= F2FS_NODE(src
);
286 struct f2fs_node
*dst_rn
= F2FS_NODE(dst
);
287 memcpy(&dst_rn
->footer
, &src_rn
->footer
, sizeof(struct node_footer
));
290 static inline void fill_node_footer_blkaddr(struct page
*page
, block_t blkaddr
)
292 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
293 struct f2fs_node
*rn
= F2FS_NODE(page
);
294 __u64 cp_ver
= cur_cp_version(ckpt
);
296 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
297 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
299 rn
->footer
.cp_ver
= cpu_to_le64(cp_ver
);
300 rn
->footer
.next_blkaddr
= cpu_to_le32(blkaddr
);
303 static inline bool is_recoverable_dnode(struct page
*page
)
305 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
306 __u64 cp_ver
= cur_cp_version(ckpt
);
308 /* Don't care crc part, if fsck.f2fs sets it. */
309 if (__is_set_ckpt_flags(ckpt
, CP_NOCRC_RECOVERY_FLAG
))
310 return (cp_ver
<< 32) == (cpver_of_node(page
) << 32);
312 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
313 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
315 return cp_ver
== cpver_of_node(page
);
319 * f2fs assigns the following node offsets described as (num).
325 * |- indirect node (3)
326 * | `- direct node (4 => 4 + N - 1)
327 * |- indirect node (4 + N)
328 * | `- direct node (5 + N => 5 + 2N - 1)
329 * `- double indirect node (5 + 2N)
330 * `- indirect node (6 + 2N)
333 * `- indirect node ((6 + 2N) + x(N + 1))
336 * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
339 static inline bool IS_DNODE(struct page
*node_page
)
341 unsigned int ofs
= ofs_of_node(node_page
);
343 if (f2fs_has_xattr_block(ofs
))
346 if (ofs
== 3 || ofs
== 4 + NIDS_PER_BLOCK
||
347 ofs
== 5 + 2 * NIDS_PER_BLOCK
)
349 if (ofs
>= 6 + 2 * NIDS_PER_BLOCK
) {
350 ofs
-= 6 + 2 * NIDS_PER_BLOCK
;
351 if (!((long int)ofs
% (NIDS_PER_BLOCK
+ 1)))
357 static inline int set_nid(struct page
*p
, int off
, nid_t nid
, bool i
)
359 struct f2fs_node
*rn
= F2FS_NODE(p
);
361 f2fs_wait_on_page_writeback(p
, NODE
, true);
364 rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
] = cpu_to_le32(nid
);
366 rn
->in
.nid
[off
] = cpu_to_le32(nid
);
367 return set_page_dirty(p
);
370 static inline nid_t
get_nid(struct page
*p
, int off
, bool i
)
372 struct f2fs_node
*rn
= F2FS_NODE(p
);
375 return le32_to_cpu(rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
]);
376 return le32_to_cpu(rn
->in
.nid
[off
]);
380 * Coldness identification:
381 * - Mark cold files in f2fs_inode_info
382 * - Mark cold node blocks in their node footer
383 * - Mark cold data pages in page cache
385 static inline int is_cold_data(struct page
*page
)
387 return PageChecked(page
);
390 static inline void set_cold_data(struct page
*page
)
392 SetPageChecked(page
);
395 static inline void clear_cold_data(struct page
*page
)
397 ClearPageChecked(page
);
400 static inline int is_node(struct page
*page
, int type
)
402 struct f2fs_node
*rn
= F2FS_NODE(page
);
403 return le32_to_cpu(rn
->footer
.flag
) & (1 << type
);
406 #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
407 #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
408 #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
410 static inline int is_inline_node(struct page
*page
)
412 return PageChecked(page
);
415 static inline void set_inline_node(struct page
*page
)
417 SetPageChecked(page
);
420 static inline void clear_inline_node(struct page
*page
)
422 ClearPageChecked(page
);
425 static inline void set_cold_node(struct inode
*inode
, struct page
*page
)
427 struct f2fs_node
*rn
= F2FS_NODE(page
);
428 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
430 if (S_ISDIR(inode
->i_mode
))
431 flag
&= ~(0x1 << COLD_BIT_SHIFT
);
433 flag
|= (0x1 << COLD_BIT_SHIFT
);
434 rn
->footer
.flag
= cpu_to_le32(flag
);
437 static inline void set_mark(struct page
*page
, int mark
, int type
)
439 struct f2fs_node
*rn
= F2FS_NODE(page
);
440 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
442 flag
|= (0x1 << type
);
444 flag
&= ~(0x1 << type
);
445 rn
->footer
.flag
= cpu_to_le32(flag
);
447 #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
448 #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)