4 * Definitions for the WAL record format.
6 * Portions Copyright (c) 1996-2015, PostgreSQL Global Development Group
7 * Portions Copyright (c) 1994, Regents of the University of California
9 * src/include/access/xlogrecord.h
14 #include "access/rmgr.h"
15 #include "access/xlogdefs.h"
16 #include "port/pg_crc32c.h"
17 #include "storage/block.h"
18 #include "storage/relfilenode.h"
21 * The overall layout of an XLOG record is:
22 * Fixed-size header (XLogRecord struct)
23 * XLogRecordBlockHeader struct
24 * XLogRecordBlockHeader struct
26 * XLogRecordDataHeader[Short|Long] struct
32 * There can be zero or more XLogRecordBlockHeaders, and 0 or more bytes of
33 * rmgr-specific data not associated with a block. XLogRecord structs
34 * always start on MAXALIGN boundaries in the WAL files, but the rest of
35 * the fields are not aligned.
37 * The XLogRecordBlockHeader, XLogRecordDataHeaderShort and
38 * XLogRecordDataHeaderLong structs all begin with a single 'id' byte. It's
39 * used to distinguish between block references, and the main data structs.
41 typedef struct XLogRecord
43 uint32 xl_tot_len; /* total len of entire record */
44 TransactionId xl_xid; /* xact id */
45 XLogRecPtr xl_prev; /* ptr to previous record in log */
46 uint8 xl_info; /* flag bits, see below */
47 RmgrId xl_rmid; /* resource manager for this record */
48 /* 2 bytes of padding here, initialize to zero */
49 pg_crc32c xl_crc; /* CRC for this record */
51 /* XLogRecordBlockHeaders and XLogRecordDataHeader follow, no padding */
55 #define SizeOfXLogRecord (offsetof(XLogRecord, xl_crc) + sizeof(pg_crc32c))
58 * The high 4 bits in xl_info may be used freely by rmgr. The
59 * XLR_SPECIAL_REL_UPDATE bit can be passed by XLogInsert caller. The rest
60 * are set internally by XLogInsert.
62 #define XLR_INFO_MASK 0x0F
63 #define XLR_RMGR_INFO_MASK 0xF0
66 * If a WAL record modifies any relation files, in ways not covered by the
67 * usual block references, this flag is set. This is not used for anything
68 * by PostgreSQL itself, but it allows external tools that read WAL and keep
69 * track of modified blocks to recognize such special record types.
71 #define XLR_SPECIAL_REL_UPDATE 0x01
74 * Header info for block data appended to an XLOG record.
76 * 'data_length' is the length of the rmgr-specific payload data associated
77 * with this block. It does not include the possible full page image, nor
78 * XLogRecordBlockHeader struct itself.
80 * Note that we don't attempt to align the XLogRecordBlockHeader struct!
81 * So, the struct must be copied to aligned local storage before use.
83 typedef struct XLogRecordBlockHeader
85 uint8 id; /* block reference ID */
86 uint8 fork_flags; /* fork within the relation, and flags */
87 uint16 data_length; /* number of payload bytes (not including page
90 /* If BKPBLOCK_HAS_IMAGE, an XLogRecordBlockImageHeader struct follows */
91 /* If BKPBLOCK_SAME_REL is not set, a RelFileNode follows */
92 /* BlockNumber follows */
93 } XLogRecordBlockHeader;
95 #define SizeOfXLogRecordBlockHeader (offsetof(XLogRecordBlockHeader, data_length) + sizeof(uint16))
98 * Additional header information when a full-page image is included
99 * (i.e. when BKPBLOCK_HAS_IMAGE is set).
101 * As a trivial form of data compression, the XLOG code is aware that
102 * PG data pages usually contain an unused "hole" in the middle, which
103 * contains only zero bytes. If the length of "hole" > 0 then we have removed
104 * such a "hole" from the stored data (and it's not counted in the
105 * XLOG record's CRC, either). Hence, the amount of block data actually
106 * present is BLCKSZ - the length of "hole" bytes.
108 * When wal_compression is enabled, a full page image which "hole" was
109 * removed is additionally compressed using PGLZ compression algorithm.
110 * This can reduce the WAL volume, but at some extra cost of CPU spent
111 * on the compression during WAL logging. In this case, since the "hole"
112 * length cannot be calculated by subtracting the number of page image bytes
113 * from BLCKSZ, basically it needs to be stored as an extra information.
114 * But when no "hole" exists, we can assume that the "hole" length is zero
115 * and no such an extra information needs to be stored. Note that
116 * the original version of page image is stored in WAL instead of the
117 * compressed one if the number of bytes saved by compression is less than
118 * the length of extra information. Hence, when a page image is successfully
119 * compressed, the amount of block data actually present is less than
120 * BLCKSZ - the length of "hole" bytes - the length of extra information.
122 typedef struct XLogRecordBlockImageHeader
124 uint16 length; /* number of page image bytes */
125 uint16 hole_offset; /* number of bytes before "hole" */
126 uint8 bimg_info; /* flag bits, see below */
129 * If BKPIMAGE_HAS_HOLE and BKPIMAGE_IS_COMPRESSED, an
130 * XLogRecordBlockCompressHeader struct follows.
132 } XLogRecordBlockImageHeader;
134 #define SizeOfXLogRecordBlockImageHeader \
135 (offsetof(XLogRecordBlockImageHeader, bimg_info) + sizeof(uint8))
137 /* Information stored in bimg_info */
138 #define BKPIMAGE_HAS_HOLE 0x01 /* page image has "hole" */
139 #define BKPIMAGE_IS_COMPRESSED 0x02 /* page image is compressed */
142 * Extra header information used when page image has "hole" and
145 typedef struct XLogRecordBlockCompressHeader
147 uint16 hole_length; /* number of bytes in "hole" */
148 } XLogRecordBlockCompressHeader;
150 #define SizeOfXLogRecordBlockCompressHeader \
151 sizeof(XLogRecordBlockCompressHeader)
154 * Maximum size of the header for a block reference. This is used to size a
155 * temporary buffer for constructing the header.
157 #define MaxSizeOfXLogRecordBlockHeader \
158 (SizeOfXLogRecordBlockHeader + \
159 SizeOfXLogRecordBlockImageHeader + \
160 SizeOfXLogRecordBlockCompressHeader + \
161 sizeof(RelFileNode) + \
165 * The fork number fits in the lower 4 bits in the fork_flags field. The upper
166 * bits are used for flags.
168 #define BKPBLOCK_FORK_MASK 0x0F
169 #define BKPBLOCK_FLAG_MASK 0xF0
170 #define BKPBLOCK_HAS_IMAGE 0x10 /* block data is an XLogRecordBlockImage */
171 #define BKPBLOCK_HAS_DATA 0x20
172 #define BKPBLOCK_WILL_INIT 0x40 /* redo will re-init the page */
173 #define BKPBLOCK_SAME_REL 0x80 /* RelFileNode omitted, same as previous */
176 * XLogRecordDataHeaderShort/Long are used for the "main data" portion of
177 * the record. If the length of the data is less than 256 bytes, the short
178 * form is used, with a single byte to hold the length. Otherwise the long
181 * (These structs are currently not used in the code, they are here just for
182 * documentation purposes).
184 typedef struct XLogRecordDataHeaderShort
186 uint8 id; /* XLR_BLOCK_ID_DATA_SHORT */
187 uint8 data_length; /* number of payload bytes */
188 } XLogRecordDataHeaderShort;
190 #define SizeOfXLogRecordDataHeaderShort (sizeof(uint8) * 2)
192 typedef struct XLogRecordDataHeaderLong
194 uint8 id; /* XLR_BLOCK_ID_DATA_LONG */
195 /* followed by uint32 data_length, unaligned */
196 } XLogRecordDataHeaderLong;
198 #define SizeOfXLogRecordDataHeaderLong (sizeof(uint8) + sizeof(uint32))
201 * Block IDs used to distinguish different kinds of record fragments. Block
202 * references are numbered from 0 to XLR_MAX_BLOCK_ID. A rmgr is free to use
203 * any ID number in that range (although you should stick to small numbers,
204 * because the WAL machinery is optimized for that case). A couple of ID
205 * numbers are reserved to denote the "main" data portion of the record.
207 * The maximum is currently set at 32, quite arbitrarily. Most records only
208 * need a handful of block references, but there are a few exceptions that
211 #define XLR_MAX_BLOCK_ID 32
213 #define XLR_BLOCK_ID_DATA_SHORT 255
214 #define XLR_BLOCK_ID_DATA_LONG 254
215 #define XLR_BLOCK_ID_ORIGIN 253
217 #endif /* XLOGRECORD_H */