1 // SPDX-License-Identifier: GPL-2.0 <<
2 /* 1 /*
3 * Copyright (c) 2003-2006, Cluster File Syste 2 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
4 * Written by Alex Tomas <alex@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-
5 */ 17 */
6 18
7 #ifndef _EXT4_EXTENTS 19 #ifndef _EXT4_EXTENTS
8 #define _EXT4_EXTENTS 20 #define _EXT4_EXTENTS
9 21
10 #include "ext4.h" 22 #include "ext4.h"
11 23
12 /* 24 /*
13 * With AGGRESSIVE_TEST defined, the capacity 25 * With AGGRESSIVE_TEST defined, the capacity of index/leaf blocks
14 * becomes very small, so index split, in-dept 26 * becomes very small, so index split, in-depth growing and
15 * other hard changes happen much more often. 27 * other hard changes happen much more often.
16 * This is for debug purposes only. 28 * This is for debug purposes only.
17 */ 29 */
18 #define AGGRESSIVE_TEST_ 30 #define AGGRESSIVE_TEST_
19 31
20 /* 32 /*
21 * With EXTENTS_STATS defined, the number of b 33 * With EXTENTS_STATS defined, the number of blocks and extents
22 * are collected in the truncate path. They'll 34 * are collected in the truncate path. They'll be shown at
23 * umount time. 35 * umount time.
24 */ 36 */
25 #define EXTENTS_STATS__ 37 #define EXTENTS_STATS__
26 38
27 /* 39 /*
28 * If CHECK_BINSEARCH is defined, then the res 40 * If CHECK_BINSEARCH is defined, then the results of the binary search
29 * will also be checked by linear search. 41 * will also be checked by linear search.
30 */ 42 */
31 #define CHECK_BINSEARCH__ 43 #define CHECK_BINSEARCH__
32 44
33 /* 45 /*
34 * If EXT_STATS is defined then stats numbers 46 * If EXT_STATS is defined then stats numbers are collected.
35 * These number will be displayed at umount ti 47 * These number will be displayed at umount time.
36 */ 48 */
37 #define EXT_STATS_ 49 #define EXT_STATS_
38 50
39 51
40 /* 52 /*
41 * ext4_inode has i_block array (60 bytes tota 53 * ext4_inode has i_block array (60 bytes total).
42 * The first 12 bytes store ext4_extent_header 54 * The first 12 bytes store ext4_extent_header;
43 * the remainder stores an array of ext4_exten 55 * the remainder stores an array of ext4_extent.
44 * For non-inode extent blocks, ext4_extent_ta 56 * For non-inode extent blocks, ext4_extent_tail
45 * follows the array. 57 * follows the array.
46 */ 58 */
47 59
48 /* 60 /*
49 * This is the extent tail on-disk structure. 61 * This is the extent tail on-disk structure.
50 * All other extent structures are 12 bytes lo 62 * All other extent structures are 12 bytes long. It turns out that
51 * block_size % 12 >= 4 for at least all power 63 * block_size % 12 >= 4 for at least all powers of 2 greater than 512, which
52 * covers all valid ext4 block sizes. Therefo 64 * covers all valid ext4 block sizes. Therefore, this tail structure can be
53 * crammed into the end of the block without h 65 * crammed into the end of the block without having to rebalance the tree.
54 */ 66 */
55 struct ext4_extent_tail { 67 struct ext4_extent_tail {
56 __le32 et_checksum; /* crc32c(uuid 68 __le32 et_checksum; /* crc32c(uuid+inum+extent_block) */
57 }; 69 };
58 70
59 /* 71 /*
60 * This is the extent on-disk structure. 72 * This is the extent on-disk structure.
61 * It's used at the bottom of the tree. 73 * It's used at the bottom of the tree.
62 */ 74 */
63 struct ext4_extent { 75 struct ext4_extent {
64 __le32 ee_block; /* first logic 76 __le32 ee_block; /* first logical block extent covers */
65 __le16 ee_len; /* number of b 77 __le16 ee_len; /* number of blocks covered by extent */
66 __le16 ee_start_hi; /* high 16 bit 78 __le16 ee_start_hi; /* high 16 bits of physical block */
67 __le32 ee_start_lo; /* low 32 bits 79 __le32 ee_start_lo; /* low 32 bits of physical block */
68 }; 80 };
69 81
70 /* 82 /*
71 * This is index on-disk structure. 83 * This is index on-disk structure.
72 * It's used at all the levels except the bott 84 * It's used at all the levels except the bottom.
73 */ 85 */
74 struct ext4_extent_idx { 86 struct ext4_extent_idx {
75 __le32 ei_block; /* index cover 87 __le32 ei_block; /* index covers logical blocks from 'block' */
76 __le32 ei_leaf_lo; /* pointer to 88 __le32 ei_leaf_lo; /* pointer to the physical block of the next *
77 * level. leaf 89 * level. leaf or next index could be there */
78 __le16 ei_leaf_hi; /* high 16 bit 90 __le16 ei_leaf_hi; /* high 16 bits of physical block */
79 __u16 ei_unused; 91 __u16 ei_unused;
80 }; 92 };
81 93
82 /* 94 /*
83 * Each block (leaves and indexes), even inode 95 * Each block (leaves and indexes), even inode-stored has header.
84 */ 96 */
85 struct ext4_extent_header { 97 struct ext4_extent_header {
86 __le16 eh_magic; /* probably wi 98 __le16 eh_magic; /* probably will support different formats */
87 __le16 eh_entries; /* number of v 99 __le16 eh_entries; /* number of valid entries */
88 __le16 eh_max; /* capacity of 100 __le16 eh_max; /* capacity of store in entries */
89 __le16 eh_depth; /* has tree re 101 __le16 eh_depth; /* has tree real underlying blocks? */
90 __le32 eh_generation; /* generation 102 __le32 eh_generation; /* generation of the tree */
91 }; 103 };
92 104
93 #define EXT4_EXT_MAGIC cpu_to_le16(0x 105 #define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
94 #define EXT4_MAX_EXTENT_DEPTH 5 106 #define EXT4_MAX_EXTENT_DEPTH 5
95 107
96 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \ 108 #define EXT4_EXTENT_TAIL_OFFSET(hdr) \
97 (sizeof(struct ext4_extent_header) + \ 109 (sizeof(struct ext4_extent_header) + \
98 (sizeof(struct ext4_extent) * le16_to 110 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
99 111
100 static inline struct ext4_extent_tail * 112 static inline struct ext4_extent_tail *
101 find_ext4_extent_tail(struct ext4_extent_heade 113 find_ext4_extent_tail(struct ext4_extent_header *eh)
102 { 114 {
103 return (struct ext4_extent_tail *)(((v 115 return (struct ext4_extent_tail *)(((void *)eh) +
104 EXT 116 EXT4_EXTENT_TAIL_OFFSET(eh));
105 } 117 }
106 118
107 /* 119 /*
108 * Array of ext4_ext_path contains path to som 120 * Array of ext4_ext_path contains path to some extent.
109 * Creation/lookup routines use it for travers 121 * Creation/lookup routines use it for traversal/splitting/etc.
110 * Truncate uses it to simulate recursive walk 122 * Truncate uses it to simulate recursive walking.
111 */ 123 */
112 struct ext4_ext_path { 124 struct ext4_ext_path {
113 ext4_fsblk_t p_bloc 125 ext4_fsblk_t p_block;
114 __u16 p_dept 126 __u16 p_depth;
115 __u16 p_maxd 127 __u16 p_maxdepth;
116 struct ext4_extent *p_ext 128 struct ext4_extent *p_ext;
117 struct ext4_extent_idx *p_idx 129 struct ext4_extent_idx *p_idx;
118 struct ext4_extent_header *p_hdr 130 struct ext4_extent_header *p_hdr;
119 struct buffer_head *p_bh; 131 struct buffer_head *p_bh;
120 }; 132 };
121 133
122 /* 134 /*
123 * Used to record a portion of a cluster found <<
124 * of an extent while traversing the extent tr <<
125 * A partial cluster may be removed if it does <<
126 * with extents that aren't being deleted (tof <<
127 * it cannot be removed (nofree state). <<
128 */ <<
129 struct partial_cluster { <<
130 ext4_fsblk_t pclu; /* physical cluste <<
131 ext4_lblk_t lblk; /* logical block n <<
132 enum {initial, tofree, nofree} state; <<
133 }; <<
134 <<
135 /* <<
136 * structure for external API 135 * structure for external API
137 */ 136 */
138 137
139 /* 138 /*
140 * EXT_INIT_MAX_LEN is the maximum number of b 139 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
141 * initialized extent. This is 2^15 and not (2 140 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
142 * MSB of ee_len field in the extent datastruc 141 * MSB of ee_len field in the extent datastructure to signify if this
143 * particular extent is an initialized extent 142 * particular extent is an initialized extent or an unwritten (i.e.
144 * preallocated). 143 * preallocated).
145 * EXT_UNWRITTEN_MAX_LEN is the maximum number 144 * EXT_UNWRITTEN_MAX_LEN is the maximum number of blocks we can have in an
146 * unwritten extent. 145 * unwritten extent.
147 * If ee_len is <= 0x8000, it is an initialize 146 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
148 * unwritten one. In other words, if MSB of ee 147 * unwritten one. In other words, if MSB of ee_len is set, it is an
149 * unwritten extent with only one special scen 148 * unwritten extent with only one special scenario when ee_len = 0x8000.
150 * In this case we can not have an unwritten e 149 * In this case we can not have an unwritten extent of zero length and
151 * thus we make it as a special case of initia 150 * thus we make it as a special case of initialized extent with 0x8000 length.
152 * This way we get better extent-to-group alig 151 * This way we get better extent-to-group alignment for initialized extents.
153 * Hence, the maximum number of blocks we can 152 * Hence, the maximum number of blocks we can have in an *initialized*
154 * extent is 2^15 (32768) and in an *unwritten 153 * extent is 2^15 (32768) and in an *unwritten* extent is 2^15-1 (32767).
155 */ 154 */
156 #define EXT_INIT_MAX_LEN (1UL << 15) 155 #define EXT_INIT_MAX_LEN (1UL << 15)
157 #define EXT_UNWRITTEN_MAX_LEN (EXT_INIT_MAX_ 156 #define EXT_UNWRITTEN_MAX_LEN (EXT_INIT_MAX_LEN - 1)
158 157
159 158
160 #define EXT_FIRST_EXTENT(__hdr__) \ 159 #define EXT_FIRST_EXTENT(__hdr__) \
161 ((struct ext4_extent *) (((char *) (__ 160 ((struct ext4_extent *) (((char *) (__hdr__)) + \
162 sizeof(struct 161 sizeof(struct ext4_extent_header)))
163 #define EXT_FIRST_INDEX(__hdr__) \ 162 #define EXT_FIRST_INDEX(__hdr__) \
164 ((struct ext4_extent_idx *) (((char *) 163 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
165 sizeof(st 164 sizeof(struct ext4_extent_header)))
166 #define EXT_HAS_FREE_INDEX(__path__) \ 165 #define EXT_HAS_FREE_INDEX(__path__) \
167 (le16_to_cpu((__path__)->p_hdr->eh_ent 166 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
168 < le16_to 167 < le16_to_cpu((__path__)->p_hdr->eh_max))
169 #define EXT_LAST_EXTENT(__hdr__) \ 168 #define EXT_LAST_EXTENT(__hdr__) \
170 (EXT_FIRST_EXTENT((__hdr__)) + le16_to 169 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
171 #define EXT_LAST_INDEX(__hdr__) \ 170 #define EXT_LAST_INDEX(__hdr__) \
172 (EXT_FIRST_INDEX((__hdr__)) + le16_to_ 171 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
173 #define EXT_MAX_EXTENT(__hdr__) \ 172 #define EXT_MAX_EXTENT(__hdr__) \
174 ((le16_to_cpu((__hdr__)->eh_max)) ? \ 173 ((le16_to_cpu((__hdr__)->eh_max)) ? \
175 ((EXT_FIRST_EXTENT((__hdr__)) + le16_t 174 ((EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)) \
176 : NULL !! 175 : 0)
177 #define EXT_MAX_INDEX(__hdr__) \ 176 #define EXT_MAX_INDEX(__hdr__) \
178 ((le16_to_cpu((__hdr__)->eh_max)) ? \ 177 ((le16_to_cpu((__hdr__)->eh_max)) ? \
179 ((EXT_FIRST_INDEX((__hdr__)) + le16_to !! 178 ((EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)) : 0)
180 : NULL <<
181 179
182 static inline struct ext4_extent_header *ext_i 180 static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
183 { 181 {
184 return (struct ext4_extent_header *) E 182 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
185 } 183 }
186 184
187 static inline struct ext4_extent_header *ext_b 185 static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
188 { 186 {
189 return (struct ext4_extent_header *) b 187 return (struct ext4_extent_header *) bh->b_data;
190 } 188 }
191 189
192 static inline unsigned short ext_depth(struct 190 static inline unsigned short ext_depth(struct inode *inode)
193 { 191 {
194 return le16_to_cpu(ext_inode_hdr(inode 192 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
195 } 193 }
196 194
197 static inline void ext4_ext_mark_unwritten(str 195 static inline void ext4_ext_mark_unwritten(struct ext4_extent *ext)
198 { 196 {
199 /* We can not have an unwritten extent 197 /* We can not have an unwritten extent of zero length! */
200 BUG_ON((le16_to_cpu(ext->ee_len) & ~EX 198 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
201 ext->ee_len |= cpu_to_le16(EXT_INIT_MA 199 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
202 } 200 }
203 201
204 static inline int ext4_ext_is_unwritten(struct 202 static inline int ext4_ext_is_unwritten(struct ext4_extent *ext)
205 { 203 {
206 /* Extent with ee_len of 0x8000 is tre 204 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
207 return (le16_to_cpu(ext->ee_len) > EXT 205 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
208 } 206 }
209 207
210 static inline int ext4_ext_get_actual_len(stru 208 static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
211 { 209 {
212 return (le16_to_cpu(ext->ee_len) <= EX 210 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
213 le16_to_cpu(ext->ee_len) : 211 le16_to_cpu(ext->ee_len) :
214 (le16_to_cpu(ext->ee_len) - EX 212 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
215 } 213 }
216 214
217 static inline void ext4_ext_mark_initialized(s 215 static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
218 { 216 {
219 ext->ee_len = cpu_to_le16(ext4_ext_get 217 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
220 } 218 }
221 219
222 /* 220 /*
223 * ext4_ext_pblock: 221 * ext4_ext_pblock:
224 * combine low and high parts of physical bloc 222 * combine low and high parts of physical block number into ext4_fsblk_t
225 */ 223 */
226 static inline ext4_fsblk_t ext4_ext_pblock(str 224 static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
227 { 225 {
228 ext4_fsblk_t block; 226 ext4_fsblk_t block;
229 227
230 block = le32_to_cpu(ex->ee_start_lo); 228 block = le32_to_cpu(ex->ee_start_lo);
231 block |= ((ext4_fsblk_t) le16_to_cpu(e 229 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
232 return block; 230 return block;
233 } 231 }
234 232
235 /* 233 /*
236 * ext4_idx_pblock: 234 * ext4_idx_pblock:
237 * combine low and high parts of a leaf physic 235 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
238 */ 236 */
239 static inline ext4_fsblk_t ext4_idx_pblock(str 237 static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
240 { 238 {
241 ext4_fsblk_t block; 239 ext4_fsblk_t block;
242 240
243 block = le32_to_cpu(ix->ei_leaf_lo); 241 block = le32_to_cpu(ix->ei_leaf_lo);
244 block |= ((ext4_fsblk_t) le16_to_cpu(i 242 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
245 return block; 243 return block;
246 } 244 }
247 245
248 /* 246 /*
249 * ext4_ext_store_pblock: 247 * ext4_ext_store_pblock:
250 * stores a large physical block number into a 248 * stores a large physical block number into an extent struct,
251 * breaking it into parts 249 * breaking it into parts
252 */ 250 */
253 static inline void ext4_ext_store_pblock(struc 251 static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
254 ext4_ 252 ext4_fsblk_t pb)
255 { 253 {
256 ex->ee_start_lo = cpu_to_le32((unsigne 254 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
257 ex->ee_start_hi = cpu_to_le16((unsigne 255 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
258 0xffff); 256 0xffff);
259 } 257 }
260 258
261 /* 259 /*
262 * ext4_idx_store_pblock: 260 * ext4_idx_store_pblock:
263 * stores a large physical block number into a 261 * stores a large physical block number into an index struct,
264 * breaking it into parts 262 * breaking it into parts
265 */ 263 */
266 static inline void ext4_idx_store_pblock(struc 264 static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
267 ext4_ 265 ext4_fsblk_t pb)
268 { 266 {
269 ix->ei_leaf_lo = cpu_to_le32((unsigned 267 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
270 ix->ei_leaf_hi = cpu_to_le16((unsigned 268 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
271 0xffff); 269 0xffff);
272 } 270 }
>> 271
>> 272 #define ext4_ext_dirty(handle, inode, path) \
>> 273 __ext4_ext_dirty(__func__, __LINE__, (handle), (inode), (path))
>> 274 int __ext4_ext_dirty(const char *where, unsigned int line, handle_t *handle,
>> 275 struct inode *inode, struct ext4_ext_path *path);
273 276
274 #endif /* _EXT4_EXTENTS */ 277 #endif /* _EXT4_EXTENTS */
275 278
276 279
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.