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