1 /*
2 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
3 * Written by Alex Tomas <alex@clusterfs.com>
4 *
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.
8 *
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.
13 *
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-
17 */
18
19 #ifndef _EXT4_EXTENTS
20 #define _EXT4_EXTENTS
21
22 #include "ext4.h"
23
24 /*
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.
29 */
30 #define AGGRESSIVE_TEST_
31
32 /*
33 * With EXTENTS_STATS defined, the number of blocks and extents
34 * are collected in the truncate path. They'll be shown at
35 * umount time.
36 */
37 #define EXTENTS_STATS__
38
39 /*
40 * If CHECK_BINSEARCH is defined, then the results of the binary search
41 * will also be checked by linear search.
42 */
43 #define CHECK_BINSEARCH__
44
45 /*
46 * If EXT_STATS is defined then stats numbers are collected.
47 * These number will be displayed at umount time.
48 */
49 #define EXT_STATS_
50
51
52 /*
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
57 * follows the array.
58 */
59
60 /*
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.
66 */
67 struct ext4_extent_tail {
68 __le32 et_checksum; /* crc32c(uuid+inum+extent_block) */
69 };
70
71 /*
72 * This is the extent on-disk structure.
73 * It's used at the bottom of the tree.
74 */
75 struct ext4_extent {
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 */
80 };
81
82 /*
83 * This is index on-disk structure.
84 * It's used at all the levels except the bottom.
85 */
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 */
91 __u16 ei_unused;
92 };
93
94 /*
95 * Each block (leaves and indexes), even inode-stored has header.
96 */
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 */
103 };
104
105 #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
106
107 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \
108 (sizeof(struct ext4_extent_header) + \
109 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
110
111 static inline struct ext4_extent_tail *
find_ext4_extent_tail(struct ext4_extent_header * eh)112 find_ext4_extent_tail(struct ext4_extent_header *eh)
113 {
114 return (struct ext4_extent_tail *)(((void *)eh) +
115 EXT4_EXTENT_TAIL_OFFSET(eh));
116 }
117
118 /*
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.
122 */
123 struct ext4_ext_path {
124 ext4_fsblk_t p_block;
125 __u16 p_depth;
126 __u16 p_maxdepth;
127 struct ext4_extent *p_ext;
128 struct ext4_extent_idx *p_idx;
129 struct ext4_extent_header *p_hdr;
130 struct buffer_head *p_bh;
131 };
132
133 /*
134 * structure for external API
135 */
136
137 /*
138 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
139 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
140 * MSB of ee_len field in the extent datastructure to signify if this
141 * particular extent is an initialized extent or an unwritten (i.e.
142 * preallocated).
143 * EXT_UNWRITTEN_MAX_LEN is the maximum number of blocks we can have in an
144 * unwritten extent.
145 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
146 * unwritten one. In other words, if MSB of ee_len is set, it is an
147 * unwritten extent with only one special scenario when ee_len = 0x8000.
148 * In this case we can not have an unwritten extent of zero length and
149 * thus we make it as a special case of initialized extent with 0x8000 length.
150 * This way we get better extent-to-group alignment for initialized extents.
151 * Hence, the maximum number of blocks we can have in an *initialized*
152 * extent is 2^15 (32768) and in an *unwritten* extent is 2^15-1 (32767).
153 */
154 #define EXT_INIT_MAX_LEN (1UL << 15)
155 #define EXT_UNWRITTEN_MAX_LEN (EXT_INIT_MAX_LEN - 1)
156
157
158 #define EXT_FIRST_EXTENT(__hdr__) \
159 ((struct ext4_extent *) (((char *) (__hdr__)) + \
160 sizeof(struct ext4_extent_header)))
161 #define EXT_FIRST_INDEX(__hdr__) \
162 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
163 sizeof(struct ext4_extent_header)))
164 #define EXT_HAS_FREE_INDEX(__path__) \
165 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
166 < le16_to_cpu((__path__)->p_hdr->eh_max))
167 #define EXT_LAST_EXTENT(__hdr__) \
168 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
169 #define EXT_LAST_INDEX(__hdr__) \
170 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
171 #define EXT_MAX_EXTENT(__hdr__) \
172 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
173 #define EXT_MAX_INDEX(__hdr__) \
174 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
175
ext_inode_hdr(struct inode * inode)176 static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
177 {
178 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
179 }
180
ext_block_hdr(struct buffer_head * bh)181 static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
182 {
183 return (struct ext4_extent_header *) bh->b_data;
184 }
185
ext_depth(struct inode * inode)186 static inline unsigned short ext_depth(struct inode *inode)
187 {
188 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
189 }
190
ext4_ext_mark_unwritten(struct ext4_extent * ext)191 static inline void ext4_ext_mark_unwritten(struct ext4_extent *ext)
192 {
193 /* We can not have an unwritten extent of zero length! */
194 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
195 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
196 }
197
ext4_ext_is_unwritten(struct ext4_extent * ext)198 static inline int ext4_ext_is_unwritten(struct ext4_extent *ext)
199 {
200 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
201 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
202 }
203
ext4_ext_get_actual_len(struct ext4_extent * ext)204 static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
205 {
206 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
207 le16_to_cpu(ext->ee_len) :
208 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
209 }
210
ext4_ext_mark_initialized(struct ext4_extent * ext)211 static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
212 {
213 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
214 }
215
216 /*
217 * ext4_ext_pblock:
218 * combine low and high parts of physical block number into ext4_fsblk_t
219 */
ext4_ext_pblock(struct ext4_extent * ex)220 static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
221 {
222 ext4_fsblk_t block;
223
224 block = le32_to_cpu(ex->ee_start_lo);
225 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
226 return block;
227 }
228
229 /*
230 * ext4_idx_pblock:
231 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
232 */
ext4_idx_pblock(struct ext4_extent_idx * ix)233 static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
234 {
235 ext4_fsblk_t block;
236
237 block = le32_to_cpu(ix->ei_leaf_lo);
238 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
239 return block;
240 }
241
242 /*
243 * ext4_ext_store_pblock:
244 * stores a large physical block number into an extent struct,
245 * breaking it into parts
246 */
ext4_ext_store_pblock(struct ext4_extent * ex,ext4_fsblk_t pb)247 static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
248 ext4_fsblk_t pb)
249 {
250 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
251 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
252 0xffff);
253 }
254
255 /*
256 * ext4_idx_store_pblock:
257 * stores a large physical block number into an index struct,
258 * breaking it into parts
259 */
ext4_idx_store_pblock(struct ext4_extent_idx * ix,ext4_fsblk_t pb)260 static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
261 ext4_fsblk_t pb)
262 {
263 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
264 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
265 0xffff);
266 }
267
268 #define ext4_ext_dirty(handle, inode, path) \
269 __ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path))
270 int __ext4_ext_dirty(const char *where, unsigned int line, handle_t *handle,
271 struct inode *inode, struct ext4_ext_path *path);
272
273 #endif /* _EXT4_EXTENTS */
274
275