src/include/access/xlogrecord.h

   1 /*
   2  * xlogrecord.h
   3  *
   4  * Definitions for the WAL record format.
   5  *
   6  * Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
   7  * Portions Copyright (c) 1994, Regents of the University of California
   8  *
   9  * src/include/access/xlogrecord.h
  10  */
  11 #ifndef XLOGRECORD_H
  12 #define XLOGRECORD_H
  13
  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"
  19
  20 /*
  21  * The overall layout of an XLOG record is:
  22  *              Fixed-size header (XLogRecord struct)
  23  *              XLogRecordBlockHeader struct
  24  *              XLogRecordBlockHeader struct
  25  *              ...
  26  *              XLogRecordDataHeader[Short|Long] struct
  27  *              block data
  28  *              block data
  29  *              ...
  30  *              main data
  31  *
  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.
  36  *
  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.
  40  */
  41 typedef struct XLogRecord
  42 {
  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 */
  50
  51         /* XLogRecordBlockHeaders and XLogRecordDataHeader follow, no padding */
  52
  53 } XLogRecord;
  54
  55 #define SizeOfXLogRecord        (offsetof(XLogRecord, xl_crc) + sizeof(pg_crc32c))
  56
  57 /*
  58  * The high 4 bits in xl_info may be used freely by rmgr. The
  59  * XLR_SPECIAL_REL_UPDATE and XLR_CHECK_CONSISTENCY bits can be passed by
  60  * XLogInsert caller. The rest are set internally by XLogInsert.
  61  */
  62 #define XLR_INFO_MASK                   0x0F
  63 #define XLR_RMGR_INFO_MASK              0xF0
  64
  65 /*
  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.
  70  */
  71 #define XLR_SPECIAL_REL_UPDATE  0x01
  72
  73 /*
  74  * Enforces consistency checks of replayed WAL at recovery. If enabled,
  75  * each record will log a full-page write for each block modified by the
  76  * record and will reuse it afterwards for consistency checks. The caller
  77  * of XLogInsert can use this value if necessary, but if
  78  * wal_consistency_checking is enabled for a rmgr this is set unconditionally.
  79  */
  80 #define XLR_CHECK_CONSISTENCY   0x02
  81
  82 /*
  83  * Header info for block data appended to an XLOG record.
  84  *
  85  * 'data_length' is the length of the rmgr-specific payload data associated
  86  * with this block. It does not include the possible full page image, nor
  87  * XLogRecordBlockHeader struct itself.
  88  *
  89  * Note that we don't attempt to align the XLogRecordBlockHeader struct!
  90  * So, the struct must be copied to aligned local storage before use.
  91  */
  92 typedef struct XLogRecordBlockHeader
  93 {
  94         uint8           id;                             /* block reference ID */
  95         uint8           fork_flags;             /* fork within the relation, and flags */
  96         uint16          data_length;    /* number of payload bytes (not including page
  97                                                                  * image) */
  98
  99         /* If BKPBLOCK_HAS_IMAGE, an XLogRecordBlockImageHeader struct follows */
 100         /* If BKPBLOCK_SAME_REL is not set, a RelFileNode follows */
 101         /* BlockNumber follows */
 102 } XLogRecordBlockHeader;
 103
 104 #define SizeOfXLogRecordBlockHeader (offsetof(XLogRecordBlockHeader, data_length) + sizeof(uint16))
 105
 106 /*
 107  * Additional header information when a full-page image is included
 108  * (i.e. when BKPBLOCK_HAS_IMAGE is set).
 109  *
 110  * The XLOG code is aware that PG data pages usually contain an unused "hole"
 111  * in the middle, which contains only zero bytes.  Since we know that the
 112  * "hole" is all zeros, we remove it from the stored data (and it's not counted
 113  * in the XLOG record's CRC, either).  Hence, the amount of block data actually
 114  * present is (BLCKSZ - <length of "hole" bytes>).
 115  *
 116  * Additionally, when wal_compression is enabled, we will try to compress full
 117  * page images using one of the supported algorithms, after removing the
 118  * "hole". This can reduce the WAL volume, but at some extra cost of CPU spent
 119  * on the compression during WAL logging. In this case, since the "hole"
 120  * length cannot be calculated by subtracting the number of page image bytes
 121  * from BLCKSZ, basically it needs to be stored as an extra information.
 122  * But when no "hole" exists, we can assume that the "hole" length is zero
 123  * and no such an extra information needs to be stored. Note that
 124  * the original version of page image is stored in WAL instead of the
 125  * compressed one if the number of bytes saved by compression is less than
 126  * the length of extra information. Hence, when a page image is successfully
 127  * compressed, the amount of block data actually present is less than
 128  * BLCKSZ - the length of "hole" bytes - the length of extra information.
 129  */
 130 typedef struct XLogRecordBlockImageHeader
 131 {
 132         uint16          length;                 /* number of page image bytes */
 133         uint16          hole_offset;    /* number of bytes before "hole" */
 134         uint8           bimg_info;              /* flag bits, see below */
 135
 136         /*
 137          * If BKPIMAGE_HAS_HOLE and BKPIMAGE_COMPRESSED(), an
 138          * XLogRecordBlockCompressHeader struct follows.
 139          */
 140 } XLogRecordBlockImageHeader;
 141
 142 #define SizeOfXLogRecordBlockImageHeader        \
 143         (offsetof(XLogRecordBlockImageHeader, bimg_info) + sizeof(uint8))
 144
 145 /* Information stored in bimg_info */
 146 #define BKPIMAGE_HAS_HOLE               0x01    /* page image has "hole" */
 147 #define BKPIMAGE_APPLY                  0x02    /* page image should be restored
 148                                                                                  * during replay */
 149 /* compression methods supported */
 150 #define BKPIMAGE_COMPRESS_PGLZ  0x04
 151 #define BKPIMAGE_COMPRESS_LZ4   0x08
 152 #define BKPIMAGE_COMPRESSED(info) \
 153         ((info & (BKPIMAGE_COMPRESS_PGLZ | BKPIMAGE_COMPRESS_LZ4)) != 0)
 154
 155 /*
 156  * Extra header information used when page image has "hole" and
 157  * is compressed.
 158  */
 159 typedef struct XLogRecordBlockCompressHeader
 160 {
 161         uint16          hole_length;    /* number of bytes in "hole" */
 162 } XLogRecordBlockCompressHeader;
 163
 164 #define SizeOfXLogRecordBlockCompressHeader \
 165         sizeof(XLogRecordBlockCompressHeader)
 166
 167 /*
 168  * Maximum size of the header for a block reference. This is used to size a
 169  * temporary buffer for constructing the header.
 170  */
 171 #define MaxSizeOfXLogRecordBlockHeader \
 172         (SizeOfXLogRecordBlockHeader + \
 173          SizeOfXLogRecordBlockImageHeader + \
 174          SizeOfXLogRecordBlockCompressHeader + \
 175          sizeof(RelFileNode) + \
 176          sizeof(BlockNumber))
 177
 178 /*
 179  * The fork number fits in the lower 4 bits in the fork_flags field. The upper
 180  * bits are used for flags.
 181  */
 182 #define BKPBLOCK_FORK_MASK      0x0F
 183 #define BKPBLOCK_FLAG_MASK      0xF0
 184 #define BKPBLOCK_HAS_IMAGE      0x10    /* block data is an XLogRecordBlockImage */
 185 #define BKPBLOCK_HAS_DATA       0x20
 186 #define BKPBLOCK_WILL_INIT      0x40    /* redo will re-init the page */
 187 #define BKPBLOCK_SAME_REL       0x80    /* RelFileNode omitted, same as previous */
 188
 189 /*
 190  * XLogRecordDataHeaderShort/Long are used for the "main data" portion of
 191  * the record. If the length of the data is less than 256 bytes, the short
 192  * form is used, with a single byte to hold the length. Otherwise the long
 193  * form is used.
 194  *
 195  * (These structs are currently not used in the code, they are here just for
 196  * documentation purposes).
 197  */
 198 typedef struct XLogRecordDataHeaderShort
 199 {
 200         uint8           id;                             /* XLR_BLOCK_ID_DATA_SHORT */
 201         uint8           data_length;    /* number of payload bytes */
 202 }                       XLogRecordDataHeaderShort;
 203
 204 #define SizeOfXLogRecordDataHeaderShort (sizeof(uint8) * 2)
 205
 206 typedef struct XLogRecordDataHeaderLong
 207 {
 208         uint8           id;                             /* XLR_BLOCK_ID_DATA_LONG */
 209         /* followed by uint32 data_length, unaligned */
 210 }                       XLogRecordDataHeaderLong;
 211
 212 #define SizeOfXLogRecordDataHeaderLong (sizeof(uint8) + sizeof(uint32))
 213
 214 /*
 215  * Block IDs used to distinguish different kinds of record fragments. Block
 216  * references are numbered from 0 to XLR_MAX_BLOCK_ID. A rmgr is free to use
 217  * any ID number in that range (although you should stick to small numbers,
 218  * because the WAL machinery is optimized for that case). A couple of ID
 219  * numbers are reserved to denote the "main" data portion of the record.
 220  *
 221  * The maximum is currently set at 32, quite arbitrarily. Most records only
 222  * need a handful of block references, but there are a few exceptions that
 223  * need more.
 224  */
 225 #define XLR_MAX_BLOCK_ID                        32
 226
 227 #define XLR_BLOCK_ID_DATA_SHORT         255
 228 #define XLR_BLOCK_ID_DATA_LONG          254
 229 #define XLR_BLOCK_ID_ORIGIN                     253
 230 #define XLR_BLOCK_ID_TOPLEVEL_XID       252
 231
 232 #endif                                                  /* XLOGRECORD_H */