1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 * http://www.samsung.com/
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)
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)
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)
18 #define DEF_RA_NID_PAGES 0 /* # of nid pages to be readaheaded */
20 /* maximum readahead size for node during getting data blocks */
21 #define MAX_RA_NODE 128
23 /* control the memory footprint threshold (10MB per 1GB ram) */
24 #define DEF_RAM_THRESHOLD 1
26 /* control dirty nats ratio threshold (default: 10% over max nid count) */
27 #define DEF_DIRTY_NAT_RATIO_THRESHOLD 10
28 /* control total # of nats */
29 #define DEF_NAT_CACHE_THRESHOLD 100000
31 /* vector size for gang look-up from nat cache that consists of radix tree */
32 #define NATVEC_SIZE 64
33 #define SETVEC_SIZE 32
35 /* return value for read_node_page */
38 /* For flag in struct node_info */
40 IS_CHECKPOINTED
, /* is it checkpointed before? */
41 HAS_FSYNCED_INODE
, /* is the inode fsynced before? */
42 HAS_LAST_FSYNC
, /* has the latest node fsync mark? */
43 IS_DIRTY
, /* this nat entry is dirty? */
44 IS_PREALLOC
, /* nat entry is preallocated */
48 * For node information
51 nid_t nid
; /* node id */
52 nid_t ino
; /* inode number of the node's owner */
53 block_t blk_addr
; /* block address of the node */
54 unsigned char version
; /* version of the node */
55 unsigned char flag
; /* for node information bits */
59 struct list_head list
; /* for clean or dirty nat list */
60 struct node_info ni
; /* in-memory node information */
63 #define nat_get_nid(nat) ((nat)->ni.nid)
64 #define nat_set_nid(nat, n) ((nat)->ni.nid = (n))
65 #define nat_get_blkaddr(nat) ((nat)->ni.blk_addr)
66 #define nat_set_blkaddr(nat, b) ((nat)->ni.blk_addr = (b))
67 #define nat_get_ino(nat) ((nat)->ni.ino)
68 #define nat_set_ino(nat, i) ((nat)->ni.ino = (i))
69 #define nat_get_version(nat) ((nat)->ni.version)
70 #define nat_set_version(nat, v) ((nat)->ni.version = (v))
72 #define inc_node_version(version) (++(version))
74 static inline void copy_node_info(struct node_info
*dst
,
75 struct node_info
*src
)
79 dst
->blk_addr
= src
->blk_addr
;
80 dst
->version
= src
->version
;
81 /* should not copy flag here */
84 static inline void set_nat_flag(struct nat_entry
*ne
,
85 unsigned int type
, bool set
)
87 unsigned char mask
= 0x01 << type
;
94 static inline bool get_nat_flag(struct nat_entry
*ne
, unsigned int type
)
96 unsigned char mask
= 0x01 << type
;
97 return ne
->ni
.flag
& mask
;
100 static inline void nat_reset_flag(struct nat_entry
*ne
)
102 /* these states can be set only after checkpoint was done */
103 set_nat_flag(ne
, IS_CHECKPOINTED
, true);
104 set_nat_flag(ne
, HAS_FSYNCED_INODE
, false);
105 set_nat_flag(ne
, HAS_LAST_FSYNC
, true);
108 static inline void node_info_from_raw_nat(struct node_info
*ni
,
109 struct f2fs_nat_entry
*raw_ne
)
111 ni
->ino
= le32_to_cpu(raw_ne
->ino
);
112 ni
->blk_addr
= le32_to_cpu(raw_ne
->block_addr
);
113 ni
->version
= raw_ne
->version
;
116 static inline void raw_nat_from_node_info(struct f2fs_nat_entry
*raw_ne
,
117 struct node_info
*ni
)
119 raw_ne
->ino
= cpu_to_le32(ni
->ino
);
120 raw_ne
->block_addr
= cpu_to_le32(ni
->blk_addr
);
121 raw_ne
->version
= ni
->version
;
124 static inline bool excess_dirty_nats(struct f2fs_sb_info
*sbi
)
126 return NM_I(sbi
)->dirty_nat_cnt
>= NM_I(sbi
)->max_nid
*
127 NM_I(sbi
)->dirty_nats_ratio
/ 100;
130 static inline bool excess_cached_nats(struct f2fs_sb_info
*sbi
)
132 return NM_I(sbi
)->nat_cnt
>= DEF_NAT_CACHE_THRESHOLD
;
135 static inline bool excess_dirty_nodes(struct f2fs_sb_info
*sbi
)
137 return get_pages(sbi
, F2FS_DIRTY_NODES
) >= sbi
->blocks_per_seg
* 8;
141 FREE_NIDS
, /* indicates the free nid list */
142 NAT_ENTRIES
, /* indicates the cached nat entry */
143 DIRTY_DENTS
, /* indicates dirty dentry pages */
144 INO_ENTRIES
, /* indicates inode entries */
145 EXTENT_CACHE
, /* indicates extent cache */
146 INMEM_PAGES
, /* indicates inmemory pages */
147 BASE_CHECK
, /* check kernel status */
150 struct nat_entry_set
{
151 struct list_head set_list
; /* link with other nat sets */
152 struct list_head entry_list
; /* link with dirty nat entries */
153 nid_t set
; /* set number*/
154 unsigned int entry_cnt
; /* the # of nat entries in set */
158 struct list_head list
; /* for free node id list */
159 nid_t nid
; /* node id */
160 int state
; /* in use or not: FREE_NID or PREALLOC_NID */
163 static inline void next_free_nid(struct f2fs_sb_info
*sbi
, nid_t
*nid
)
165 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
166 struct free_nid
*fnid
;
168 spin_lock(&nm_i
->nid_list_lock
);
169 if (nm_i
->nid_cnt
[FREE_NID
] <= 0) {
170 spin_unlock(&nm_i
->nid_list_lock
);
173 fnid
= list_first_entry(&nm_i
->free_nid_list
, struct free_nid
, list
);
175 spin_unlock(&nm_i
->nid_list_lock
);
181 static inline void get_nat_bitmap(struct f2fs_sb_info
*sbi
, void *addr
)
183 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
185 #ifdef CONFIG_F2FS_CHECK_FS
186 if (memcmp(nm_i
->nat_bitmap
, nm_i
->nat_bitmap_mir
,
190 memcpy(addr
, nm_i
->nat_bitmap
, nm_i
->bitmap_size
);
193 static inline pgoff_t
current_nat_addr(struct f2fs_sb_info
*sbi
, nid_t start
)
195 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
200 * block_off = segment_off * 512 + off_in_segment
201 * OLD = (segment_off * 512) * 2 + off_in_segment
202 * NEW = 2 * (segment_off * 512 + off_in_segment) - off_in_segment
204 block_off
= NAT_BLOCK_OFFSET(start
);
206 block_addr
= (pgoff_t
)(nm_i
->nat_blkaddr
+
208 (block_off
& (sbi
->blocks_per_seg
- 1)));
210 if (f2fs_test_bit(block_off
, nm_i
->nat_bitmap
))
211 block_addr
+= sbi
->blocks_per_seg
;
216 static inline pgoff_t
next_nat_addr(struct f2fs_sb_info
*sbi
,
219 struct f2fs_nm_info
*nm_i
= NM_I(sbi
);
221 block_addr
-= nm_i
->nat_blkaddr
;
222 block_addr
^= 1 << sbi
->log_blocks_per_seg
;
223 return block_addr
+ nm_i
->nat_blkaddr
;
226 static inline void set_to_next_nat(struct f2fs_nm_info
*nm_i
, nid_t start_nid
)
228 unsigned int block_off
= NAT_BLOCK_OFFSET(start_nid
);
230 f2fs_change_bit(block_off
, nm_i
->nat_bitmap
);
231 #ifdef CONFIG_F2FS_CHECK_FS
232 f2fs_change_bit(block_off
, nm_i
->nat_bitmap_mir
);
236 static inline nid_t
ino_of_node(struct page
*node_page
)
238 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
239 return le32_to_cpu(rn
->footer
.ino
);
242 static inline nid_t
nid_of_node(struct page
*node_page
)
244 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
245 return le32_to_cpu(rn
->footer
.nid
);
248 static inline unsigned int ofs_of_node(struct page
*node_page
)
250 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
251 unsigned flag
= le32_to_cpu(rn
->footer
.flag
);
252 return flag
>> OFFSET_BIT_SHIFT
;
255 static inline __u64
cpver_of_node(struct page
*node_page
)
257 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
258 return le64_to_cpu(rn
->footer
.cp_ver
);
261 static inline block_t
next_blkaddr_of_node(struct page
*node_page
)
263 struct f2fs_node
*rn
= F2FS_NODE(node_page
);
264 return le32_to_cpu(rn
->footer
.next_blkaddr
);
267 static inline void fill_node_footer(struct page
*page
, nid_t nid
,
268 nid_t ino
, unsigned int ofs
, bool reset
)
270 struct f2fs_node
*rn
= F2FS_NODE(page
);
271 unsigned int old_flag
= 0;
274 memset(rn
, 0, sizeof(*rn
));
276 old_flag
= le32_to_cpu(rn
->footer
.flag
);
278 rn
->footer
.nid
= cpu_to_le32(nid
);
279 rn
->footer
.ino
= cpu_to_le32(ino
);
281 /* should remain old flag bits such as COLD_BIT_SHIFT */
282 rn
->footer
.flag
= cpu_to_le32((ofs
<< OFFSET_BIT_SHIFT
) |
283 (old_flag
& OFFSET_BIT_MASK
));
286 static inline void copy_node_footer(struct page
*dst
, struct page
*src
)
288 struct f2fs_node
*src_rn
= F2FS_NODE(src
);
289 struct f2fs_node
*dst_rn
= F2FS_NODE(dst
);
290 memcpy(&dst_rn
->footer
, &src_rn
->footer
, sizeof(struct node_footer
));
293 static inline void fill_node_footer_blkaddr(struct page
*page
, block_t blkaddr
)
295 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
296 struct f2fs_node
*rn
= F2FS_NODE(page
);
297 __u64 cp_ver
= cur_cp_version(ckpt
);
299 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
300 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
302 rn
->footer
.cp_ver
= cpu_to_le64(cp_ver
);
303 rn
->footer
.next_blkaddr
= cpu_to_le32(blkaddr
);
306 static inline bool is_recoverable_dnode(struct page
*page
)
308 struct f2fs_checkpoint
*ckpt
= F2FS_CKPT(F2FS_P_SB(page
));
309 __u64 cp_ver
= cur_cp_version(ckpt
);
311 /* Don't care crc part, if fsck.f2fs sets it. */
312 if (__is_set_ckpt_flags(ckpt
, CP_NOCRC_RECOVERY_FLAG
))
313 return (cp_ver
<< 32) == (cpver_of_node(page
) << 32);
315 if (__is_set_ckpt_flags(ckpt
, CP_CRC_RECOVERY_FLAG
))
316 cp_ver
|= (cur_cp_crc(ckpt
) << 32);
318 return cp_ver
== cpver_of_node(page
);
322 * f2fs assigns the following node offsets described as (num).
328 * |- indirect node (3)
329 * | `- direct node (4 => 4 + N - 1)
330 * |- indirect node (4 + N)
331 * | `- direct node (5 + N => 5 + 2N - 1)
332 * `- double indirect node (5 + 2N)
333 * `- indirect node (6 + 2N)
336 * `- indirect node ((6 + 2N) + x(N + 1))
339 * `- indirect node ((6 + 2N) + (N - 1)(N + 1))
342 static inline bool IS_DNODE(struct page
*node_page
)
344 unsigned int ofs
= ofs_of_node(node_page
);
346 if (f2fs_has_xattr_block(ofs
))
349 if (ofs
== 3 || ofs
== 4 + NIDS_PER_BLOCK
||
350 ofs
== 5 + 2 * NIDS_PER_BLOCK
)
352 if (ofs
>= 6 + 2 * NIDS_PER_BLOCK
) {
353 ofs
-= 6 + 2 * NIDS_PER_BLOCK
;
354 if (!((long int)ofs
% (NIDS_PER_BLOCK
+ 1)))
360 static inline int set_nid(struct page
*p
, int off
, nid_t nid
, bool i
)
362 struct f2fs_node
*rn
= F2FS_NODE(p
);
364 f2fs_wait_on_page_writeback(p
, NODE
, true, true);
367 rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
] = cpu_to_le32(nid
);
369 rn
->in
.nid
[off
] = cpu_to_le32(nid
);
370 return set_page_dirty(p
);
373 static inline nid_t
get_nid(struct page
*p
, int off
, bool i
)
375 struct f2fs_node
*rn
= F2FS_NODE(p
);
378 return le32_to_cpu(rn
->i
.i_nid
[off
- NODE_DIR1_BLOCK
]);
379 return le32_to_cpu(rn
->in
.nid
[off
]);
383 * Coldness identification:
384 * - Mark cold files in f2fs_inode_info
385 * - Mark cold node blocks in their node footer
386 * - Mark cold data pages in page cache
388 static inline int is_cold_data(struct page
*page
)
390 return PageChecked(page
);
393 static inline void set_cold_data(struct page
*page
)
395 SetPageChecked(page
);
398 static inline void clear_cold_data(struct page
*page
)
400 ClearPageChecked(page
);
403 static inline int is_node(struct page
*page
, int type
)
405 struct f2fs_node
*rn
= F2FS_NODE(page
);
406 return le32_to_cpu(rn
->footer
.flag
) & (1 << type
);
409 #define is_cold_node(page) is_node(page, COLD_BIT_SHIFT)
410 #define is_fsync_dnode(page) is_node(page, FSYNC_BIT_SHIFT)
411 #define is_dent_dnode(page) is_node(page, DENT_BIT_SHIFT)
413 static inline int is_inline_node(struct page
*page
)
415 return PageChecked(page
);
418 static inline void set_inline_node(struct page
*page
)
420 SetPageChecked(page
);
423 static inline void clear_inline_node(struct page
*page
)
425 ClearPageChecked(page
);
428 static inline void set_cold_node(struct page
*page
, bool is_dir
)
430 struct f2fs_node
*rn
= F2FS_NODE(page
);
431 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
434 flag
&= ~(0x1 << COLD_BIT_SHIFT
);
436 flag
|= (0x1 << COLD_BIT_SHIFT
);
437 rn
->footer
.flag
= cpu_to_le32(flag
);
440 static inline void set_mark(struct page
*page
, int mark
, int type
)
442 struct f2fs_node
*rn
= F2FS_NODE(page
);
443 unsigned int flag
= le32_to_cpu(rn
->footer
.flag
);
445 flag
|= (0x1 << type
);
447 flag
&= ~(0x1 << type
);
448 rn
->footer
.flag
= cpu_to_le32(flag
);
450 #ifdef CONFIG_F2FS_CHECK_FS
451 f2fs_inode_chksum_set(F2FS_P_SB(page
), page
);
454 #define set_dentry_mark(page, mark) set_mark(page, mark, DENT_BIT_SHIFT)
455 #define set_fsync_mark(page, mark) set_mark(page, mark, FSYNC_BIT_SHIFT)