blob: 8643ff5bbeb7a9954af0026ce43601d02487856b [file] [log] [blame]
Alex Tomasa86c6182006-10-11 01:21:03 -07001/*
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
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040019#ifndef _EXT4_EXTENTS
20#define _EXT4_EXTENTS
Alex Tomasa86c6182006-10-11 01:21:03 -070021
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040022#include "ext4.h"
Alex Tomasa86c6182006-10-11 01:21:03 -070023
24/*
Robert P. J. Daybbf2f9f2007-02-17 19:20:16 +010025 * With AGGRESSIVE_TEST defined, the capacity of index/leaf blocks
Randy Dunlapd0d856e2006-10-11 01:21:07 -070026 * 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.
Alex Tomasa86c6182006-10-11 01:21:03 -070029 */
Robert P. J. Daybbf2f9f2007-02-17 19:20:16 +010030#define AGGRESSIVE_TEST_
Alex Tomasa86c6182006-10-11 01:21:03 -070031
32/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070033 * 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.
Alex Tomasa86c6182006-10-11 01:21:03 -070036 */
37#define EXTENTS_STATS__
38
39/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070040 * If CHECK_BINSEARCH is defined, then the results of the binary search
41 * will also be checked by linear search.
Alex Tomasa86c6182006-10-11 01:21:03 -070042 */
43#define CHECK_BINSEARCH__
44
45/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070046 * If EXT_STATS is defined then stats numbers are collected.
47 * These number will be displayed at umount time.
Alex Tomasa86c6182006-10-11 01:21:03 -070048 */
49#define EXT_STATS_
50
51
52/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070053 * 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.
Darrick J. Wonge6153912012-04-29 18:23:10 -040056 * For non-inode extent blocks, ext4_extent_tail
57 * follows the array.
Alex Tomasa86c6182006-10-11 01:21:03 -070058 */
59
60/*
Darrick J. Wonge6153912012-04-29 18:23:10 -040061 * 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 */
67struct ext4_extent_tail {
68 __le32 et_checksum; /* crc32c(uuid+inum+extent_block) */
69};
70
71/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070072 * This is the extent on-disk structure.
73 * It's used at the bottom of the tree.
Alex Tomasa86c6182006-10-11 01:21:03 -070074 */
75struct 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 */
Aneesh Kumar K.Vb3776112007-10-16 18:38:25 -040079 __le32 ee_start_lo; /* low 32 bits of physical block */
Alex Tomasa86c6182006-10-11 01:21:03 -070080};
81
82/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070083 * This is index on-disk structure.
84 * It's used at all the levels except the bottom.
Alex Tomasa86c6182006-10-11 01:21:03 -070085 */
86struct ext4_extent_idx {
87 __le32 ei_block; /* index covers logical blocks from 'block' */
Aneesh Kumar K.Vd8dd0b42007-10-16 18:38:25 -040088 __le32 ei_leaf_lo; /* pointer to the physical block of the next *
Randy Dunlapd0d856e2006-10-11 01:21:07 -070089 * level. leaf or next index could be there */
Alex Tomasa86c6182006-10-11 01:21:03 -070090 __le16 ei_leaf_hi; /* high 16 bits of physical block */
91 __u16 ei_unused;
92};
93
94/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -070095 * Each block (leaves and indexes), even inode-stored has header.
Alex Tomasa86c6182006-10-11 01:21:03 -070096 */
97struct 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 */
Randy Dunlapd0d856e2006-10-11 01:21:07 -0700101 __le16 eh_depth; /* has tree real underlying blocks? */
Alex Tomasa86c6182006-10-11 01:21:03 -0700102 __le32 eh_generation; /* generation of the tree */
103};
104
105#define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
106
Darrick J. Wong7ac59902012-04-29 18:37:10 -0400107#define EXT4_EXTENT_TAIL_OFFSET(hdr) \
108 (sizeof(struct ext4_extent_header) + \
109 (sizeof(struct ext4_extent) * le16_to_cpu((hdr)->eh_max)))
110
111static inline struct ext4_extent_tail *
112find_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
Alex Tomasa86c6182006-10-11 01:21:03 -0700118/*
Randy Dunlapd0d856e2006-10-11 01:21:07 -0700119 * 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.
Alex Tomasa86c6182006-10-11 01:21:03 -0700122 */
123struct ext4_ext_path {
Alex Tomasf65e6fb2006-10-11 01:21:05 -0700124 ext4_fsblk_t p_block;
Alex Tomasa86c6182006-10-11 01:21:03 -0700125 __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
Eric Sandeen6873fa02008-10-07 00:46:36 -0400136/*
Lukas Czernerf17722f2011-06-06 00:05:17 -0400137 * Maximum number of logical blocks in a file; ext4_extent's ee_block is
138 * __le32.
139 */
140#define EXT_MAX_BLOCKS 0xffffffff
Alex Tomasa86c6182006-10-11 01:21:03 -0700141
Amit Arora749269f2007-07-18 09:02:56 -0400142/*
143 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
144 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
145 * MSB of ee_len field in the extent datastructure to signify if this
146 * particular extent is an initialized extent or an uninitialized (i.e.
147 * preallocated).
148 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
149 * uninitialized extent.
150 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
151 * uninitialized one. In other words, if MSB of ee_len is set, it is an
152 * uninitialized extent with only one special scenario when ee_len = 0x8000.
153 * In this case we can not have an uninitialized extent of zero length and
154 * thus we make it as a special case of initialized extent with 0x8000 length.
155 * This way we get better extent-to-group alignment for initialized extents.
156 * Hence, the maximum number of blocks we can have in an *initialized*
157 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
158 */
159#define EXT_INIT_MAX_LEN (1UL << 15)
160#define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
Suparna Bhattacharya471d4012006-10-11 01:21:06 -0700161
Alex Tomasa86c6182006-10-11 01:21:03 -0700162
163#define EXT_FIRST_EXTENT(__hdr__) \
164 ((struct ext4_extent *) (((char *) (__hdr__)) + \
165 sizeof(struct ext4_extent_header)))
166#define EXT_FIRST_INDEX(__hdr__) \
167 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
168 sizeof(struct ext4_extent_header)))
169#define EXT_HAS_FREE_INDEX(__path__) \
Dave Kleikamp8c55e202007-05-24 13:04:54 -0400170 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
171 < le16_to_cpu((__path__)->p_hdr->eh_max))
Alex Tomasa86c6182006-10-11 01:21:03 -0700172#define EXT_LAST_EXTENT(__hdr__) \
173 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
174#define EXT_LAST_INDEX(__hdr__) \
175 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
176#define EXT_MAX_EXTENT(__hdr__) \
177 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
178#define EXT_MAX_INDEX(__hdr__) \
179 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
180
181static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
182{
183 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
184}
185
186static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
187{
188 return (struct ext4_extent_header *) bh->b_data;
189}
190
191static inline unsigned short ext_depth(struct inode *inode)
192{
193 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
194}
195
Amit Aroraa2df2a62007-07-17 21:42:41 -0400196static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext)
197{
Amit Arora749269f2007-07-18 09:02:56 -0400198 /* We can not have an uninitialized extent of zero length! */
199 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
200 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
Amit Aroraa2df2a62007-07-17 21:42:41 -0400201}
202
203static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext)
204{
Amit Arora749269f2007-07-18 09:02:56 -0400205 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
206 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
Amit Aroraa2df2a62007-07-17 21:42:41 -0400207}
208
209static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
210{
Amit Arora749269f2007-07-18 09:02:56 -0400211 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
212 le16_to_cpu(ext->ee_len) :
213 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
Amit Aroraa2df2a62007-07-17 21:42:41 -0400214}
215
Mingming Cao00314622009-09-28 15:49:08 -0400216static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
217{
218 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
219}
220
Theodore Ts'obf89d162010-10-27 21:30:14 -0400221/*
222 * ext4_ext_pblock:
223 * combine low and high parts of physical block number into ext4_fsblk_t
224 */
225static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
226{
227 ext4_fsblk_t block;
228
229 block = le32_to_cpu(ex->ee_start_lo);
230 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
231 return block;
232}
233
234/*
235 * ext4_idx_pblock:
236 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
237 */
238static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
239{
240 ext4_fsblk_t block;
241
242 block = le32_to_cpu(ix->ei_leaf_lo);
243 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
244 return block;
245}
246
247/*
248 * ext4_ext_store_pblock:
249 * stores a large physical block number into an extent struct,
250 * breaking it into parts
251 */
252static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
253 ext4_fsblk_t pb)
254{
255 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
256 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
257 0xffff);
258}
259
260/*
261 * ext4_idx_store_pblock:
262 * stores a large physical block number into an index struct,
263 * breaking it into parts
264 */
265static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
266 ext4_fsblk_t pb)
267{
268 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
269 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
270 0xffff);
271}
272
Christoph Hellwig3dcf5452008-04-29 18:13:32 -0400273#endif /* _EXT4_EXTENTS */
Alex Tomasa86c6182006-10-11 01:21:03 -0700274