2 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
3 * Written by Alex Tomas <alex@clusterfs.com>
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
25 * With AGGRESSIVE_TEST defined, the capacity of index/leaf blocks
26 * becomes very small, so index split, in-depth growing and
27 * other hard changes happen much more often.
28 * This is for debug purposes only.
30 #define AGGRESSIVE_TEST_
33 * With EXTENTS_STATS defined, the number of blocks and extents
34 * are collected in the truncate path. They'll be shown at
37 #define EXTENTS_STATS__
40 * If CHECK_BINSEARCH is defined, then the results of the binary search
41 * will also be checked by linear search.
43 #define CHECK_BINSEARCH__
46 * If EXT_STATS is defined then stats numbers are collected.
47 * These number will be displayed at umount time.
53 * ext4_inode has i_block array (60 bytes total).
54 * The first 12 bytes store ext4_extent_header;
55 * the remainder stores an array of ext4_extent.
56 * For non-inode extent blocks, ext4_extent_tail
61 * This is the extent tail on-disk structure.
62 * All other extent structures are 12 bytes long. It turns out that
63 * block_size % 12 >= 4 for at least all powers of 2 greater than 512, which
64 * covers all valid ext4 block sizes. Therefore, this tail structure can be
65 * crammed into the end of the block without having to rebalance the tree.
67 struct ext4_extent_tail
{
68 __le32 et_checksum
; /* crc32c(uuid+inum+extent_block) */
72 * This is the extent on-disk structure.
73 * It's used at the bottom of the tree.
76 __le32 ee_block
; /* first logical block extent covers */
77 __le16 ee_len
; /* number of blocks covered by extent */
78 __le16 ee_start_hi
; /* high 16 bits of physical block */
79 __le32 ee_start_lo
; /* low 32 bits of physical block */
83 * This is index on-disk structure.
84 * It's used at all the levels except the bottom.
86 struct ext4_extent_idx
{
87 __le32 ei_block
; /* index covers logical blocks from 'block' */
88 __le32 ei_leaf_lo
; /* pointer to the physical block of the next *
89 * level. leaf or next index could be there */
90 __le16 ei_leaf_hi
; /* high 16 bits of physical block */
95 * Each block (leaves and indexes), even inode-stored has header.
97 struct ext4_extent_header
{
98 __le16 eh_magic
; /* probably will support different formats */
99 __le16 eh_entries
; /* number of valid entries */
100 __le16 eh_max
; /* capacity of store in entries */
101 __le16 eh_depth
; /* has tree real underlying blocks? */
102 __le32 eh_generation
; /* generation of the tree */
105 #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
106 #define EXT4_MAX_EXTENT_DEPTH 5
108 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \
109 (sizeof(struct ext4_extent_header) + \
110 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
112 static inline struct ext4_extent_tail
*
113 find_ext4_extent_tail(struct ext4_extent_header
*eh
)
115 return (struct ext4_extent_tail
*)(((void *)eh
) +
116 EXT4_EXTENT_TAIL_OFFSET(eh
));
120 * Array of ext4_ext_path contains path to some extent.
121 * Creation/lookup routines use it for traversal/splitting/etc.
122 * Truncate uses it to simulate recursive walking.
124 struct ext4_ext_path
{
125 ext4_fsblk_t p_block
;
128 struct ext4_extent
*p_ext
;
129 struct ext4_extent_idx
*p_idx
;
130 struct ext4_extent_header
*p_hdr
;
131 struct buffer_head
*p_bh
;
135 * structure for external API
139 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
140 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
141 * MSB of ee_len field in the extent datastructure to signify if this
142 * particular extent is an initialized extent or an unwritten (i.e.
144 * EXT_UNWRITTEN_MAX_LEN is the maximum number of blocks we can have in an
146 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
147 * unwritten one. In other words, if MSB of ee_len is set, it is an
148 * unwritten extent with only one special scenario when ee_len = 0x8000.
149 * In this case we can not have an unwritten extent of zero length and
150 * thus we make it as a special case of initialized extent with 0x8000 length.
151 * This way we get better extent-to-group alignment for initialized extents.
152 * Hence, the maximum number of blocks we can have in an *initialized*
153 * extent is 2^15 (32768) and in an *unwritten* extent is 2^15-1 (32767).
155 #define EXT_INIT_MAX_LEN (1UL << 15)
156 #define EXT_UNWRITTEN_MAX_LEN (EXT_INIT_MAX_LEN - 1)
159 #define EXT_FIRST_EXTENT(__hdr__) \
160 ((struct ext4_extent *) (((char *) (__hdr__)) + \
161 sizeof(struct ext4_extent_header)))
162 #define EXT_FIRST_INDEX(__hdr__) \
163 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
164 sizeof(struct ext4_extent_header)))
165 #define EXT_HAS_FREE_INDEX(__path__) \
166 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
167 < le16_to_cpu((__path__)->p_hdr->eh_max))
168 #define EXT_LAST_EXTENT(__hdr__) \
169 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
170 #define EXT_LAST_INDEX(__hdr__) \
171 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
172 #define EXT_MAX_EXTENT(__hdr__) \
173 ((le16_to_cpu((__hdr__)->eh_max)) ? \
174 ((EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)) \
176 #define EXT_MAX_INDEX(__hdr__) \
177 ((le16_to_cpu((__hdr__)->eh_max)) ? \
178 ((EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)) : 0)
180 static inline struct ext4_extent_header
*ext_inode_hdr(struct inode
*inode
)
182 return (struct ext4_extent_header
*) EXT4_I(inode
)->i_data
;
185 static inline struct ext4_extent_header
*ext_block_hdr(struct buffer_head
*bh
)
187 return (struct ext4_extent_header
*) bh
->b_data
;
190 static inline unsigned short ext_depth(struct inode
*inode
)
192 return le16_to_cpu(ext_inode_hdr(inode
)->eh_depth
);
195 static inline void ext4_ext_mark_unwritten(struct ext4_extent
*ext
)
197 /* We can not have an unwritten extent of zero length! */
198 BUG_ON((le16_to_cpu(ext
->ee_len
) & ~EXT_INIT_MAX_LEN
) == 0);
199 ext
->ee_len
|= cpu_to_le16(EXT_INIT_MAX_LEN
);
202 static inline int ext4_ext_is_unwritten(struct ext4_extent
*ext
)
204 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
205 return (le16_to_cpu(ext
->ee_len
) > EXT_INIT_MAX_LEN
);
208 static inline int ext4_ext_get_actual_len(struct ext4_extent
*ext
)
210 return (le16_to_cpu(ext
->ee_len
) <= EXT_INIT_MAX_LEN
?
211 le16_to_cpu(ext
->ee_len
) :
212 (le16_to_cpu(ext
->ee_len
) - EXT_INIT_MAX_LEN
));
215 static inline void ext4_ext_mark_initialized(struct ext4_extent
*ext
)
217 ext
->ee_len
= cpu_to_le16(ext4_ext_get_actual_len(ext
));
222 * combine low and high parts of physical block number into ext4_fsblk_t
224 static inline ext4_fsblk_t
ext4_ext_pblock(struct ext4_extent
*ex
)
228 block
= le32_to_cpu(ex
->ee_start_lo
);
229 block
|= ((ext4_fsblk_t
) le16_to_cpu(ex
->ee_start_hi
) << 31) << 1;
235 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
237 static inline ext4_fsblk_t
ext4_idx_pblock(struct ext4_extent_idx
*ix
)
241 block
= le32_to_cpu(ix
->ei_leaf_lo
);
242 block
|= ((ext4_fsblk_t
) le16_to_cpu(ix
->ei_leaf_hi
) << 31) << 1;
247 * ext4_ext_store_pblock:
248 * stores a large physical block number into an extent struct,
249 * breaking it into parts
251 static inline void ext4_ext_store_pblock(struct ext4_extent
*ex
,
254 ex
->ee_start_lo
= cpu_to_le32((unsigned long) (pb
& 0xffffffff));
255 ex
->ee_start_hi
= cpu_to_le16((unsigned long) ((pb
>> 31) >> 1) &
260 * ext4_idx_store_pblock:
261 * stores a large physical block number into an index struct,
262 * breaking it into parts
264 static inline void ext4_idx_store_pblock(struct ext4_extent_idx
*ix
,
267 ix
->ei_leaf_lo
= cpu_to_le32((unsigned long) (pb
& 0xffffffff));
268 ix
->ei_leaf_hi
= cpu_to_le16((unsigned long) ((pb
>> 31) >> 1) &
272 #define ext4_ext_dirty(handle, inode, path) \
273 __ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path))
274 int __ext4_ext_dirty(const char *where
, unsigned int line
, handle_t
*handle
,
275 struct inode
*inode
, struct ext4_ext_path
*path
);
277 #endif /* _EXT4_EXTENTS */