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-
19 #ifndef _LINUX_EXT4_EXTENTS
20 #define _LINUX_EXT4_EXTENTS
22 #include <linux/ext4_fs.h>
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_DEBUG is defined you can use the 'extdebug' mount option
47 * to get lots of info about what's going on.
51 #define ext_debug(a...) printk(a)
53 #define ext_debug(a...)
57 * If EXT_STATS is defined then stats numbers are collected.
58 * These number will be displayed at umount time.
64 * ext4_inode has i_block array (60 bytes total).
65 * The first 12 bytes store ext4_extent_header;
66 * the remainder stores an array of ext4_extent.
70 * This is the extent on-disk structure.
71 * It's used at the bottom of the tree.
74 __le32 ee_block; /* first logical block extent covers */
75 __le16 ee_len; /* number of blocks covered by extent */
76 __le16 ee_start_hi; /* high 16 bits of physical block */
77 __le32 ee_start; /* low 32 bits of physical block */
81 * This is index on-disk structure.
82 * It's used at all the levels except the bottom.
84 struct ext4_extent_idx {
85 __le32 ei_block; /* index covers logical blocks from 'block' */
86 __le32 ei_leaf; /* pointer to the physical block of the next *
87 * level. leaf or next index could be there */
88 __le16 ei_leaf_hi; /* high 16 bits of physical block */
93 * Each block (leaves and indexes), even inode-stored has header.
95 struct ext4_extent_header {
96 __le16 eh_magic; /* probably will support different formats */
97 __le16 eh_entries; /* number of valid entries */
98 __le16 eh_max; /* capacity of store in entries */
99 __le16 eh_depth; /* has tree real underlying blocks? */
100 __le32 eh_generation; /* generation of the tree */
103 #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
106 * Array of ext4_ext_path contains path to some extent.
107 * Creation/lookup routines use it for traversal/splitting/etc.
108 * Truncate uses it to simulate recursive walking.
110 struct ext4_ext_path {
111 ext4_fsblk_t p_block;
113 struct ext4_extent *p_ext;
114 struct ext4_extent_idx *p_idx;
115 struct ext4_extent_header *p_hdr;
116 struct buffer_head *p_bh;
120 * structure for external API
123 #define EXT4_EXT_CACHE_NO 0
124 #define EXT4_EXT_CACHE_GAP 1
125 #define EXT4_EXT_CACHE_EXTENT 2
128 * to be called by ext4_ext_walk_space()
129 * negative retcode - error
130 * positive retcode - signal for ext4_ext_walk_space(), see below
131 * callback must return valid extent (passed or newly created)
133 typedef int (*ext_prepare_callback)(struct inode *, struct ext4_ext_path *,
134 struct ext4_ext_cache *,
137 #define EXT_CONTINUE 0
142 #define EXT_MAX_BLOCK 0xffffffff
145 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
146 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
147 * MSB of ee_len field in the extent datastructure to signify if this
148 * particular extent is an initialized extent or an uninitialized (i.e.
150 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
151 * uninitialized extent.
152 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
153 * uninitialized one. In other words, if MSB of ee_len is set, it is an
154 * uninitialized extent with only one special scenario when ee_len = 0x8000.
155 * In this case we can not have an uninitialized extent of zero length and
156 * thus we make it as a special case of initialized extent with 0x8000 length.
157 * This way we get better extent-to-group alignment for initialized extents.
158 * Hence, the maximum number of blocks we can have in an *initialized*
159 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
161 #define EXT_INIT_MAX_LEN (1UL << 15)
162 #define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
165 #define EXT_FIRST_EXTENT(__hdr__) \
166 ((struct ext4_extent *) (((char *) (__hdr__)) + \
167 sizeof(struct ext4_extent_header)))
168 #define EXT_FIRST_INDEX(__hdr__) \
169 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
170 sizeof(struct ext4_extent_header)))
171 #define EXT_HAS_FREE_INDEX(__path__) \
172 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
173 < le16_to_cpu((__path__)->p_hdr->eh_max))
174 #define EXT_LAST_EXTENT(__hdr__) \
175 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
176 #define EXT_LAST_INDEX(__hdr__) \
177 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
178 #define EXT_MAX_EXTENT(__hdr__) \
179 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
180 #define EXT_MAX_INDEX(__hdr__) \
181 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
183 static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
185 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
188 static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
190 return (struct ext4_extent_header *) bh->b_data;
193 static inline unsigned short ext_depth(struct inode *inode)
195 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
198 static inline void ext4_ext_tree_changed(struct inode *inode)
200 EXT4_I(inode)->i_ext_generation++;
204 ext4_ext_invalidate_cache(struct inode *inode)
206 EXT4_I(inode)->i_cached_extent.ec_type = EXT4_EXT_CACHE_NO;
209 static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext)
211 /* We can not have an uninitialized extent of zero length! */
212 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
213 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
216 static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext)
218 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
219 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
222 static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
224 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
225 le16_to_cpu(ext->ee_len) :
226 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
229 extern int ext4_extent_tree_init(handle_t *, struct inode *);
230 extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *);
231 extern int ext4_ext_try_to_merge(struct inode *inode,
232 struct ext4_ext_path *path,
233 struct ext4_extent *);
234 extern unsigned int ext4_ext_check_overlap(struct inode *, struct ext4_extent *, struct ext4_ext_path *);
235 extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *);
236 extern int ext4_ext_walk_space(struct inode *, unsigned long, unsigned long, ext_prepare_callback, void *);
237 extern struct ext4_ext_path * ext4_ext_find_extent(struct inode *, int, struct ext4_ext_path *);
239 #endif /* _LINUX_EXT4_EXTENTS */