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 Licens
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)
107 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \
108 (sizeof(struct ext4_extent_header) + \
109 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
111 static inline struct ext4_extent_tail *
112 find_ext4_extent_tail(struct ext4_extent_header *eh)
114 return (struct ext4_extent_tail *)(((void *)eh) +
115 EXT4_EXTENT_TAIL_OFFSET(eh));
119 * Array of ext4_ext_path contains path to some extent.
120 * Creation/lookup routines use it for traversal/splitting/etc.
121 * Truncate uses it to simulate recursive walking.
123 struct ext4_ext_path {
124 ext4_fsblk_t p_block;
126 struct ext4_extent *p_ext;
127 struct ext4_extent_idx *p_idx;
128 struct ext4_extent_header *p_hdr;
129 struct buffer_head *p_bh;
133 * structure for external API
137 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
140 #define EXT_MAX_BLOCKS 0xffffffff
143 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
144 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
145 * MSB of ee_len field in the extent datastructure to signify if this
146 * particular extent is an initialized extent or an uninitialized (i.e.
148 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
149 * uninitialized extent.
150 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
151 * uninitialized one. In other words, if MSB of ee_len is set, it is an
152 * uninitialized extent with only one special scenario when ee_len = 0x8000.
153 * In this case we can not have an uninitialized extent of zero length and
154 * thus we make it as a special case of initialized extent with 0x8000 length.
155 * This way we get better extent-to-group alignment for initialized extents.
156 * Hence, the maximum number of blocks we can have in an *initialized*
157 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
159 #define EXT_INIT_MAX_LEN (1UL << 15)
160 #define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
163 #define EXT_FIRST_EXTENT(__hdr__) \
164 ((struct ext4_extent *) (((char *) (__hdr__)) + \
165 sizeof(struct ext4_extent_header)))
166 #define EXT_FIRST_INDEX(__hdr__) \
167 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
168 sizeof(struct ext4_extent_header)))
169 #define EXT_HAS_FREE_INDEX(__path__) \
170 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
171 < le16_to_cpu((__path__)->p_hdr->eh_max))
172 #define EXT_LAST_EXTENT(__hdr__) \
173 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
174 #define EXT_LAST_INDEX(__hdr__) \
175 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
176 #define EXT_MAX_EXTENT(__hdr__) \
177 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
178 #define EXT_MAX_INDEX(__hdr__) \
179 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
181 static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
183 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
186 static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
188 return (struct ext4_extent_header *) bh->b_data;
191 static inline unsigned short ext_depth(struct inode *inode)
193 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
196 static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext)
198 /* We can not have an uninitialized extent of zero length! */
199 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
200 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
203 static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext)
205 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
206 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
209 static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
211 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
212 le16_to_cpu(ext->ee_len) :
213 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
216 static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
218 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
223 * combine low and high parts of physical block number into ext4_fsblk_t
225 static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
229 block = le32_to_cpu(ex->ee_start_lo);
230 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
236 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
238 static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
242 block = le32_to_cpu(ix->ei_leaf_lo);
243 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
248 * ext4_ext_store_pblock:
249 * stores a large physical block number into an extent struct,
250 * breaking it into parts
252 static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
255 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
256 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
261 * ext4_idx_store_pblock:
262 * stores a large physical block number into an index struct,
263 * breaking it into parts
265 static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
268 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
269 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
273 #endif /* _EXT4_EXTENTS */