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 * Turn on EXT_DEBUG to get lots of info about extents operations.
50 #define ext_debug(fmt, ...) printk(fmt, ##__VA_ARGS__)
52 #define ext_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
56 * If EXT_STATS is defined then stats numbers are collected.
57 * These number will be displayed at umount time.
63 * ext4_inode has i_block array (60 bytes total).
64 * The first 12 bytes store ext4_extent_header;
65 * the remainder stores an array of ext4_extent.
66 * For non-inode extent blocks, ext4_extent_tail
71 * This is the extent tail on-disk structure.
72 * All other extent structures are 12 bytes long. It turns out that
73 * block_size % 12 >= 4 for at least all powers of 2 greater than 512, which
74 * covers all valid ext4 block sizes. Therefore, this tail structure can be
75 * crammed into the end of the block without having to rebalance the tree.
77 struct ext4_extent_tail {
78 __le32 et_checksum; /* crc32c(uuid+inum+extent_block) */
82 * This is the extent on-disk structure.
83 * It's used at the bottom of the tree.
86 __le32 ee_block; /* first logical block extent covers */
87 __le16 ee_len; /* number of blocks covered by extent */
88 __le16 ee_start_hi; /* high 16 bits of physical block */
89 __le32 ee_start_lo; /* low 32 bits of physical block */
93 * This is index on-disk structure.
94 * It's used at all the levels except the bottom.
96 struct ext4_extent_idx {
97 __le32 ei_block; /* index covers logical blocks from 'block' */
98 __le32 ei_leaf_lo; /* pointer to the physical block of the next *
99 * level. leaf or next index could be there */
100 __le16 ei_leaf_hi; /* high 16 bits of physical block */
105 * Each block (leaves and indexes), even inode-stored has header.
107 struct ext4_extent_header {
108 __le16 eh_magic; /* probably will support different formats */
109 __le16 eh_entries; /* number of valid entries */
110 __le16 eh_max; /* capacity of store in entries */
111 __le16 eh_depth; /* has tree real underlying blocks? */
112 __le32 eh_generation; /* generation of the tree */
115 #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
118 * Array of ext4_ext_path contains path to some extent.
119 * Creation/lookup routines use it for traversal/splitting/etc.
120 * Truncate uses it to simulate recursive walking.
122 struct ext4_ext_path {
123 ext4_fsblk_t p_block;
125 struct ext4_extent *p_ext;
126 struct ext4_extent_idx *p_idx;
127 struct ext4_extent_header *p_hdr;
128 struct buffer_head *p_bh;
132 * structure for external API
136 * to be called by ext4_ext_walk_space()
137 * negative retcode - error
138 * positive retcode - signal for ext4_ext_walk_space(), see below
139 * callback must return valid extent (passed or newly created)
141 typedef int (*ext_prepare_callback)(struct inode *, ext4_lblk_t,
142 struct ext4_ext_cache *,
143 struct ext4_extent *, void *);
145 #define EXT_CONTINUE 0
150 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
153 #define EXT_MAX_BLOCKS 0xffffffff
156 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
157 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
158 * MSB of ee_len field in the extent datastructure to signify if this
159 * particular extent is an initialized extent or an uninitialized (i.e.
161 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
162 * uninitialized extent.
163 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
164 * uninitialized one. In other words, if MSB of ee_len is set, it is an
165 * uninitialized extent with only one special scenario when ee_len = 0x8000.
166 * In this case we can not have an uninitialized extent of zero length and
167 * thus we make it as a special case of initialized extent with 0x8000 length.
168 * This way we get better extent-to-group alignment for initialized extents.
169 * Hence, the maximum number of blocks we can have in an *initialized*
170 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
172 #define EXT_INIT_MAX_LEN (1UL << 15)
173 #define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
176 #define EXT_FIRST_EXTENT(__hdr__) \
177 ((struct ext4_extent *) (((char *) (__hdr__)) + \
178 sizeof(struct ext4_extent_header)))
179 #define EXT_FIRST_INDEX(__hdr__) \
180 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
181 sizeof(struct ext4_extent_header)))
182 #define EXT_HAS_FREE_INDEX(__path__) \
183 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
184 < le16_to_cpu((__path__)->p_hdr->eh_max))
185 #define EXT_LAST_EXTENT(__hdr__) \
186 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
187 #define EXT_LAST_INDEX(__hdr__) \
188 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
189 #define EXT_MAX_EXTENT(__hdr__) \
190 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
191 #define EXT_MAX_INDEX(__hdr__) \
192 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
194 static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
196 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
199 static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
201 return (struct ext4_extent_header *) bh->b_data;
204 static inline unsigned short ext_depth(struct inode *inode)
206 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
210 ext4_ext_invalidate_cache(struct inode *inode)
212 EXT4_I(inode)->i_cached_extent.ec_len = 0;
215 static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext)
217 /* We can not have an uninitialized extent of zero length! */
218 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
219 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
222 static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext)
224 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
225 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
228 static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
230 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
231 le16_to_cpu(ext->ee_len) :
232 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
235 static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
237 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
242 * combine low and high parts of physical block number into ext4_fsblk_t
244 static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
248 block = le32_to_cpu(ex->ee_start_lo);
249 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
255 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
257 static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
261 block = le32_to_cpu(ix->ei_leaf_lo);
262 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
267 * ext4_ext_store_pblock:
268 * stores a large physical block number into an extent struct,
269 * breaking it into parts
271 static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
274 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
275 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
280 * ext4_idx_store_pblock:
281 * stores a large physical block number into an index struct,
282 * breaking it into parts
284 static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
287 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
288 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
292 extern int ext4_ext_calc_metadata_amount(struct inode *inode,
293 ext4_lblk_t lblocks);
294 extern int ext4_extent_tree_init(handle_t *, struct inode *);
295 extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
297 struct ext4_ext_path *path);
298 extern int ext4_can_extents_be_merged(struct inode *inode,
299 struct ext4_extent *ex1,
300 struct ext4_extent *ex2);
301 extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *, int);
302 extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
303 struct ext4_ext_path *);
304 extern void ext4_ext_drop_refs(struct ext4_ext_path *);
305 extern int ext4_ext_check_inode(struct inode *inode);
306 extern int ext4_find_delalloc_cluster(struct inode *inode, ext4_lblk_t lblk,
307 int search_hint_reverse);
308 #endif /* _EXT4_EXTENTS */