Commit | Line | Data |
---|---|---|
a86c6181 AT |
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 | ||
3dcf5451 CH |
19 | #ifndef _EXT4_EXTENTS |
20 | #define _EXT4_EXTENTS | |
a86c6181 | 21 | |
3dcf5451 | 22 | #include "ext4.h" |
a86c6181 AT |
23 | |
24 | /* | |
bbf2f9fb | 25 | * With AGGRESSIVE_TEST defined, the capacity of index/leaf blocks |
d0d856e8 RD |
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. | |
a86c6181 | 29 | */ |
bbf2f9fb | 30 | #define AGGRESSIVE_TEST_ |
a86c6181 AT |
31 | |
32 | /* | |
d0d856e8 RD |
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. | |
a86c6181 AT |
36 | */ |
37 | #define EXTENTS_STATS__ | |
38 | ||
39 | /* | |
d0d856e8 RD |
40 | * If CHECK_BINSEARCH is defined, then the results of the binary search |
41 | * will also be checked by linear search. | |
a86c6181 AT |
42 | */ |
43 | #define CHECK_BINSEARCH__ | |
44 | ||
a86c6181 | 45 | /* |
d0d856e8 RD |
46 | * If EXT_STATS is defined then stats numbers are collected. |
47 | * These number will be displayed at umount time. | |
a86c6181 AT |
48 | */ |
49 | #define EXT_STATS_ | |
50 | ||
51 | ||
52 | /* | |
d0d856e8 RD |
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. | |
e6153918 DW |
56 | * For non-inode extent blocks, ext4_extent_tail |
57 | * follows the array. | |
a86c6181 AT |
58 | */ |
59 | ||
e6153918 DW |
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 | ||
a86c6181 | 71 | /* |
d0d856e8 RD |
72 | * This is the extent on-disk structure. |
73 | * It's used at the bottom of the tree. | |
a86c6181 AT |
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 */ | |
b377611d | 79 | __le32 ee_start_lo; /* low 32 bits of physical block */ |
a86c6181 AT |
80 | }; |
81 | ||
82 | /* | |
d0d856e8 RD |
83 | * This is index on-disk structure. |
84 | * It's used at all the levels except the bottom. | |
a86c6181 AT |
85 | */ |
86 | struct ext4_extent_idx { | |
87 | __le32 ei_block; /* index covers logical blocks from 'block' */ | |
d8dd0b45 | 88 | __le32 ei_leaf_lo; /* pointer to the physical block of the next * |
d0d856e8 | 89 | * level. leaf or next index could be there */ |
a86c6181 AT |
90 | __le16 ei_leaf_hi; /* high 16 bits of physical block */ |
91 | __u16 ei_unused; | |
92 | }; | |
93 | ||
94 | /* | |
d0d856e8 | 95 | * Each block (leaves and indexes), even inode-stored has header. |
a86c6181 AT |
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 */ | |
d0d856e8 | 101 | __le16 eh_depth; /* has tree real underlying blocks? */ |
a86c6181 AT |
102 | __le32 eh_generation; /* generation of the tree */ |
103 | }; | |
104 | ||
105 | #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a) | |
106 | ||
7ac5990d DW |
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 * | |
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 | ||
a86c6181 | 118 | /* |
d0d856e8 RD |
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. | |
a86c6181 AT |
122 | */ |
123 | struct ext4_ext_path { | |
f65e6fba | 124 | ext4_fsblk_t p_block; |
a86c6181 AT |
125 | __u16 p_depth; |
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; | |
130 | }; | |
131 | ||
132 | /* | |
133 | * structure for external API | |
134 | */ | |
135 | ||
749269fa AA |
136 | /* |
137 | * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an | |
138 | * initialized extent. This is 2^15 and not (2^16 - 1), since we use the | |
139 | * MSB of ee_len field in the extent datastructure to signify if this | |
140 | * particular extent is an initialized extent or an uninitialized (i.e. | |
141 | * preallocated). | |
142 | * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an | |
143 | * uninitialized extent. | |
144 | * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an | |
145 | * uninitialized one. In other words, if MSB of ee_len is set, it is an | |
146 | * uninitialized extent with only one special scenario when ee_len = 0x8000. | |
147 | * In this case we can not have an uninitialized extent of zero length and | |
148 | * thus we make it as a special case of initialized extent with 0x8000 length. | |
149 | * This way we get better extent-to-group alignment for initialized extents. | |
150 | * Hence, the maximum number of blocks we can have in an *initialized* | |
151 | * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767). | |
152 | */ | |
153 | #define EXT_INIT_MAX_LEN (1UL << 15) | |
154 | #define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1) | |
471d4011 | 155 | |
a86c6181 AT |
156 | |
157 | #define EXT_FIRST_EXTENT(__hdr__) \ | |
158 | ((struct ext4_extent *) (((char *) (__hdr__)) + \ | |
159 | sizeof(struct ext4_extent_header))) | |
160 | #define EXT_FIRST_INDEX(__hdr__) \ | |
161 | ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \ | |
162 | sizeof(struct ext4_extent_header))) | |
163 | #define EXT_HAS_FREE_INDEX(__path__) \ | |
8c55e204 DK |
164 | (le16_to_cpu((__path__)->p_hdr->eh_entries) \ |
165 | < le16_to_cpu((__path__)->p_hdr->eh_max)) | |
a86c6181 AT |
166 | #define EXT_LAST_EXTENT(__hdr__) \ |
167 | (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1) | |
168 | #define EXT_LAST_INDEX(__hdr__) \ | |
169 | (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1) | |
170 | #define EXT_MAX_EXTENT(__hdr__) \ | |
171 | (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1) | |
172 | #define EXT_MAX_INDEX(__hdr__) \ | |
173 | (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1) | |
174 | ||
175 | static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode) | |
176 | { | |
177 | return (struct ext4_extent_header *) EXT4_I(inode)->i_data; | |
178 | } | |
179 | ||
180 | static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh) | |
181 | { | |
182 | return (struct ext4_extent_header *) bh->b_data; | |
183 | } | |
184 | ||
185 | static inline unsigned short ext_depth(struct inode *inode) | |
186 | { | |
187 | return le16_to_cpu(ext_inode_hdr(inode)->eh_depth); | |
188 | } | |
189 | ||
a2df2a63 AA |
190 | static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext) |
191 | { | |
749269fa AA |
192 | /* We can not have an uninitialized extent of zero length! */ |
193 | BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0); | |
194 | ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN); | |
a2df2a63 AA |
195 | } |
196 | ||
197 | static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext) | |
198 | { | |
749269fa AA |
199 | /* Extent with ee_len of 0x8000 is treated as an initialized extent */ |
200 | return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN); | |
a2df2a63 AA |
201 | } |
202 | ||
203 | static inline int ext4_ext_get_actual_len(struct ext4_extent *ext) | |
204 | { | |
749269fa AA |
205 | return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ? |
206 | le16_to_cpu(ext->ee_len) : | |
207 | (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN)); | |
a2df2a63 AA |
208 | } |
209 | ||
0031462b MC |
210 | static inline void ext4_ext_mark_initialized(struct ext4_extent *ext) |
211 | { | |
212 | ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext)); | |
213 | } | |
214 | ||
bf89d16f TT |
215 | /* |
216 | * ext4_ext_pblock: | |
217 | * combine low and high parts of physical block number into ext4_fsblk_t | |
218 | */ | |
219 | static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex) | |
220 | { | |
221 | ext4_fsblk_t block; | |
222 | ||
223 | block = le32_to_cpu(ex->ee_start_lo); | |
224 | block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1; | |
225 | return block; | |
226 | } | |
227 | ||
228 | /* | |
229 | * ext4_idx_pblock: | |
230 | * combine low and high parts of a leaf physical block number into ext4_fsblk_t | |
231 | */ | |
232 | static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix) | |
233 | { | |
234 | ext4_fsblk_t block; | |
235 | ||
236 | block = le32_to_cpu(ix->ei_leaf_lo); | |
237 | block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1; | |
238 | return block; | |
239 | } | |
240 | ||
241 | /* | |
242 | * ext4_ext_store_pblock: | |
243 | * stores a large physical block number into an extent struct, | |
244 | * breaking it into parts | |
245 | */ | |
246 | static inline void ext4_ext_store_pblock(struct ext4_extent *ex, | |
247 | ext4_fsblk_t pb) | |
248 | { | |
249 | ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff)); | |
250 | ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & | |
251 | 0xffff); | |
252 | } | |
253 | ||
254 | /* | |
255 | * ext4_idx_store_pblock: | |
256 | * stores a large physical block number into an index struct, | |
257 | * breaking it into parts | |
258 | */ | |
259 | static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix, | |
260 | ext4_fsblk_t pb) | |
261 | { | |
262 | ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff)); | |
263 | ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & | |
264 | 0xffff); | |
265 | } | |
266 | ||
2656497b DW |
267 | #define ext4_ext_dirty(handle, inode, path) \ |
268 | __ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path)) | |
269 | int __ext4_ext_dirty(const char *where, unsigned int line, handle_t *handle, | |
270 | struct inode *inode, struct ext4_ext_path *path); | |
271 | ||
3dcf5451 | 272 | #endif /* _EXT4_EXTENTS */ |
a86c6181 | 273 |