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)
117 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \
118 (sizeof(struct ext4_extent_header) + \
119 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
121 static inline struct ext4_extent_tail
*
122 find_ext4_extent_tail(struct ext4_extent_header
*eh
)
124 return (struct ext4_extent_tail
*)(((void *)eh
) +
125 EXT4_EXTENT_TAIL_OFFSET(eh
));
129 * Array of ext4_ext_path contains path to some extent.
130 * Creation/lookup routines use it for traversal/splitting/etc.
131 * Truncate uses it to simulate recursive walking.
133 struct ext4_ext_path
{
134 ext4_fsblk_t p_block
;
136 struct ext4_extent
*p_ext
;
137 struct ext4_extent_idx
*p_idx
;
138 struct ext4_extent_header
*p_hdr
;
139 struct buffer_head
*p_bh
;
143 * structure for external API
147 * to be called by ext4_ext_walk_space()
148 * negative retcode - error
149 * positive retcode - signal for ext4_ext_walk_space(), see below
150 * callback must return valid extent (passed or newly created)
152 typedef int (*ext_prepare_callback
)(struct inode
*, ext4_lblk_t
,
153 struct ext4_ext_cache
*,
154 struct ext4_extent
*, void *);
156 #define EXT_CONTINUE 0
161 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
164 #define EXT_MAX_BLOCKS 0xffffffff
167 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
168 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
169 * MSB of ee_len field in the extent datastructure to signify if this
170 * particular extent is an initialized extent or an uninitialized (i.e.
172 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
173 * uninitialized extent.
174 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
175 * uninitialized one. In other words, if MSB of ee_len is set, it is an
176 * uninitialized extent with only one special scenario when ee_len = 0x8000.
177 * In this case we can not have an uninitialized extent of zero length and
178 * thus we make it as a special case of initialized extent with 0x8000 length.
179 * This way we get better extent-to-group alignment for initialized extents.
180 * Hence, the maximum number of blocks we can have in an *initialized*
181 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
183 #define EXT_INIT_MAX_LEN (1UL << 15)
184 #define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
187 #define EXT_FIRST_EXTENT(__hdr__) \
188 ((struct ext4_extent *) (((char *) (__hdr__)) + \
189 sizeof(struct ext4_extent_header)))
190 #define EXT_FIRST_INDEX(__hdr__) \
191 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
192 sizeof(struct ext4_extent_header)))
193 #define EXT_HAS_FREE_INDEX(__path__) \
194 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
195 < le16_to_cpu((__path__)->p_hdr->eh_max))
196 #define EXT_LAST_EXTENT(__hdr__) \
197 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
198 #define EXT_LAST_INDEX(__hdr__) \
199 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
200 #define EXT_MAX_EXTENT(__hdr__) \
201 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
202 #define EXT_MAX_INDEX(__hdr__) \
203 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
205 static inline struct ext4_extent_header
*ext_inode_hdr(struct inode
*inode
)
207 return (struct ext4_extent_header
*) EXT4_I(inode
)->i_data
;
210 static inline struct ext4_extent_header
*ext_block_hdr(struct buffer_head
*bh
)
212 return (struct ext4_extent_header
*) bh
->b_data
;
215 static inline unsigned short ext_depth(struct inode
*inode
)
217 return le16_to_cpu(ext_inode_hdr(inode
)->eh_depth
);
221 ext4_ext_invalidate_cache(struct inode
*inode
)
223 EXT4_I(inode
)->i_cached_extent
.ec_len
= 0;
226 static inline void ext4_ext_mark_uninitialized(struct ext4_extent
*ext
)
228 /* We can not have an uninitialized extent of zero length! */
229 BUG_ON((le16_to_cpu(ext
->ee_len
) & ~EXT_INIT_MAX_LEN
) == 0);
230 ext
->ee_len
|= cpu_to_le16(EXT_INIT_MAX_LEN
);
233 static inline int ext4_ext_is_uninitialized(struct ext4_extent
*ext
)
235 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
236 return (le16_to_cpu(ext
->ee_len
) > EXT_INIT_MAX_LEN
);
239 static inline int ext4_ext_get_actual_len(struct ext4_extent
*ext
)
241 return (le16_to_cpu(ext
->ee_len
) <= EXT_INIT_MAX_LEN
?
242 le16_to_cpu(ext
->ee_len
) :
243 (le16_to_cpu(ext
->ee_len
) - EXT_INIT_MAX_LEN
));
246 static inline void ext4_ext_mark_initialized(struct ext4_extent
*ext
)
248 ext
->ee_len
= cpu_to_le16(ext4_ext_get_actual_len(ext
));
253 * combine low and high parts of physical block number into ext4_fsblk_t
255 static inline ext4_fsblk_t
ext4_ext_pblock(struct ext4_extent
*ex
)
259 block
= le32_to_cpu(ex
->ee_start_lo
);
260 block
|= ((ext4_fsblk_t
) le16_to_cpu(ex
->ee_start_hi
) << 31) << 1;
266 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
268 static inline ext4_fsblk_t
ext4_idx_pblock(struct ext4_extent_idx
*ix
)
272 block
= le32_to_cpu(ix
->ei_leaf_lo
);
273 block
|= ((ext4_fsblk_t
) le16_to_cpu(ix
->ei_leaf_hi
) << 31) << 1;
278 * ext4_ext_store_pblock:
279 * stores a large physical block number into an extent struct,
280 * breaking it into parts
282 static inline void ext4_ext_store_pblock(struct ext4_extent
*ex
,
285 ex
->ee_start_lo
= cpu_to_le32((unsigned long) (pb
& 0xffffffff));
286 ex
->ee_start_hi
= cpu_to_le16((unsigned long) ((pb
>> 31) >> 1) &
291 * ext4_idx_store_pblock:
292 * stores a large physical block number into an index struct,
293 * breaking it into parts
295 static inline void ext4_idx_store_pblock(struct ext4_extent_idx
*ix
,
298 ix
->ei_leaf_lo
= cpu_to_le32((unsigned long) (pb
& 0xffffffff));
299 ix
->ei_leaf_hi
= cpu_to_le16((unsigned long) ((pb
>> 31) >> 1) &
303 extern int ext4_ext_calc_metadata_amount(struct inode
*inode
,
304 ext4_lblk_t lblocks
);
305 extern int ext4_extent_tree_init(handle_t
*, struct inode
*);
306 extern int ext4_ext_calc_credits_for_single_extent(struct inode
*inode
,
308 struct ext4_ext_path
*path
);
309 extern int ext4_can_extents_be_merged(struct inode
*inode
,
310 struct ext4_extent
*ex1
,
311 struct ext4_extent
*ex2
);
312 extern int ext4_ext_insert_extent(handle_t
*, struct inode
*, struct ext4_ext_path
*, struct ext4_extent
*, int);
313 extern struct ext4_ext_path
*ext4_ext_find_extent(struct inode
*, ext4_lblk_t
,
314 struct ext4_ext_path
*);
315 extern void ext4_ext_drop_refs(struct ext4_ext_path
*);
316 extern int ext4_ext_check_inode(struct inode
*inode
);
317 extern int ext4_find_delalloc_cluster(struct inode
*inode
, ext4_lblk_t lblk
,
318 int search_hint_reverse
);
319 #endif /* _EXT4_EXTENTS */