PostgreSQL启动过程中的那些事七：初始化共享内存和信号二：shmem中初始化xlog

pg初始化完shmem，给其加上索引"ShmemIndex"后，接着就在shmem里初始化xlog。

1先上个图，看一下函数调用过程梗概，中间略过部分细节

初始化xlog方法调用流程图

2初始化xlog相关结构

话说main()->…->PostmasterMain()->…->reset_shared() ->CreateSharedMemoryAndSemaphores()>…->XLOGSHmemInit()，初始化控制文件data/global/pg_control相关数据结构及事务日志xlog相关数据结构，相关结构定义在下面。

typedef struct ControlFileData

{

/*

* Unique system identifier --- to ensure wematch up xlog files with the

* installation that produced them.

*/

uint64 system_identifier;

/*

* Version identifier information. Keep these fields at the same offset,

* especially pg_control_version; they won't bereal useful if they move

* around. (Forhistorical reasons they must be 8 bytes into the file

* rather than immediately at the front.)

*

* pg_control_version identifies the format ofpg_control itself.

* catalog_version_no identifies the format ofthe system catalogs.

*

* There are additional version identifiers inindividual files; for

* example, WAL logs contain per-page magic numbersthat can serve as

* version cues for the WAL log.

*/

uint32 pg_control_version; /* PG_CONTROL_VERSION */

uint32 catalog_version_no; /* see catversion.h */

/*

* System status data

*/

DBState state; /*see enum above */

pg_time_t time; /*time stamp of last pg_control update */

XLogRecPtr checkPoint; /*last check point record ptr */

XLogRecPtr prevCheckPoint; /* previous check point recordptr */

CheckPoint checkPointCopy; /* copy of last check pointrecord */

/*

* These two values determine the minimum pointwe must recover up to

* before starting up:

*

* minRecoveryPoint is updated to the latestreplayed LSN whenever we

* flush a data change during archive recovery.That guards against

* starting archive recovery, aborting it, andrestarting with an earlier

* stop location. If we've already flushed datachanges from WAL record X

* to disk, we mustn't start up until we reachX again. Zero when not

* doing archive recovery.

*

* backupStartPoint is the redo pointer of thebackup start checkpoint, if

* we are recovering from an online backup andhaven't reached the end of

* backup yet. It is reset to zero when the endof backup is reached, and

* we mustn't start up before that. A booleanwould suffice otherwise, but

* we use the redo pointer as a cross-checkwhen we see an end-of-backup

* record, to make sure the end-of-backuprecord corresponds the base

* backup we're recovering from.

*/

XLogRecPtr minRecoveryPoint;

XLogRecPtr backupStartPoint;

/*

* Parameter settings that determine if the WALcan be used for archival

* or hot standby.

*/

int wal_level;

int MaxConnections;

int max_prepared_xacts;

int max_locks_per_xact;

/*

* This data is used to check for hardware-architecturecompatibility of

* the database and the backendexecutable. We need not check endianness

* explicitly, since the pg_control versionwill surely look wrong to a

* machine of different endianness, but we doneed to worry about MAXALIGN

* and floating-point format. (Note: storage layout nominally also

* depends on SHORTALIGN and INTALIGN, but inpractice these are the same

* on all architectures of interest.)

*

* Testing just one double value is not a verybulletproof test for

* floating-point compatibility, but it willcatch most cases.

*/

uint32 maxAlign; /* alignment requirement for tuples */

double floatFormat; /* constant 1234567.0 */

#define FLOATFORMAT_VALUE 1234567.0

/*

* This data is used to make sure that configurationof this database is

* compatible with the backend executable.

*/

uint32 blcksz; /* data block size for this DB */

uint32 relseg_size; /* blocks per segment of large relation */

uint32 xlog_blcksz; /* block size within WAL files */

uint32 xlog_seg_size; /* size of each WAL segment */

uint32 nameDataLen; /* catalog name field width */

uint32 indexMaxKeys; /* max number of columns in an index */

uint32 toast_max_chunk_size; /* chunk size in TOAST tables */

/*flag indicating internal format of timestamp, interval, time */

bool enableIntTimes; /* int64 storageenabled? */

/*flags indicating pass-by-value status of various types */

bool float4ByVal; /* float4 pass-by-value? */

bool float8ByVal; /* float8, int8, etc pass-by-value? */

/*CRC of all above ... MUST BE LAST! */

pg_crc32 crc;

} ControlFileData;

/*

* Bodyof CheckPoint XLOG records. This isdeclared here because we keep

* acopy of the latest one in pg_control for possible disaster recovery.

*Changing this struct requires a PG_CONTROL_VERSION bump.

*/

typedef struct CheckPoint

{

XLogRecPtr redo; /*next RecPtr available when we began to

* create CheckPoint (i.e. REDO start point) */

TimeLineID ThisTimeLineID; /* current TLI */

uint32 nextXidEpoch; /* higher-order bits of nextXid */

TransactionIdnextXid; /* next free XID */

Oid nextOid; /* next free OID */

MultiXactIdnextMulti; /* next freeMultiXactId */

MultiXactOffsetnextMultiOffset; /* next free MultiXactoffset */

TransactionIdoldestXid; /* cluster-wide minimumdatfrozenxid */

Oid oldestXidDB; /* database with minimum datfrozenxid */

pg_time_t time; /*time stamp of checkpoint */

/*

* Oldest XID still running. This is onlyneeded to initialize hot standby

* mode from an online checkpoint, so we onlybother calculating this for

* online checkpoints and only when wal_levelis hot_standby. Otherwise

* it's set to InvalidTransactionId.

*/

TransactionIdoldestActiveXid;

} CheckPoint;

/*

* Total shared-memorystate for XLOG.

*/

typedef struct XLogCtlData

{

/* Protected byWALInsertLock: */

XLogCtlInsertInsert;

/* Protected byinfo_lck: */

XLogwrtRqstLogwrtRqst;

XLogwrtResultLogwrtResult;

uint32 ckptXidEpoch; /* nextXID & epoch of latest checkpoint */

TransactionIdckptXid;

XLogRecPtr asyncXactLSN; /*LSN of newest async commit/abort */

uint32 lastRemovedLog; /* latest removed/recycledXLOG segment */

uint32 lastRemovedSeg;

/* Protected byWALWriteLock: */

XLogCtlWrite Write;

/*

* These values do not change after startup,although the pointed-to pages

* and xlblocks values certainly do. Permission to read/write the pages

* and xlblocks values depends on WALInsertLockand WALWriteLock.

*/

char *pages; /* buffers forunwritten XLOG pages */

XLogRecPtr*xlblocks; /* 1st byte ptr-s +XLOG_BLCKSZ */

int XLogCacheBlck; /* highest allocated xlog buffer index */

TimeLineID ThisTimeLineID;

TimeLineID RecoveryTargetTLI;

/*

* archiveCleanupCommand is read fromrecovery.conf but needs to be in

* shared memory so that the bgwriter processcan access it.

*/

char archiveCleanupCommand[MAXPGPATH];

/*

* SharedRecoveryInProgress indicates if we'restill in crash or archive

* recovery.Protected by info_lck.

*/

bool SharedRecoveryInProgress;

/*

* SharedHotStandbyActive indicates if we'restill in crash or archive

* recovery.Protected by info_lck.

*/

bool SharedHotStandbyActive;

/*

* recoveryWakeupLatch is used to wake up thestartup process to continue

* WAL replay, if it is waiting for WAL toarrive or failover trigger file

* to appear.

*/

Latch recoveryWakeupLatch;

/*

* During recovery, we keep a copy of thelatest checkpoint record here.

* Used by the background writer when it wantsto create a restartpoint.

*

* Protected by info_lck.

*/

XLogRecPtr lastCheckPointRecPtr;

CheckPoint lastCheckPoint;

/* end+1 of the lastrecord replayed (or being replayed) */

XLogRecPtr replayEndRecPtr;

/* end+1 of the lastrecord replayed */

XLogRecPtr recoveryLastRecPtr;

/* timestamp of lastCOMMIT/ABORT record replayed (or being replayed) */

TimestampTzrecoveryLastXTime;

/* Are we requestedto pause recovery? */

bool recoveryPause;

slock_t info_lck; /*locks shared variables shown above */

} XLogCtlData;

/*

* Shared state datafor XLogInsert.

*/

typedef struct XLogCtlInsert

{

XLogwrtResultLogwrtResult; /* a recent value of LogwrtResult */

XLogRecPtr PrevRecord; /*start of previously-inserted record */

int curridx; /* current block index in cache */

XLogPageHeadercurrpage; /* points to header of blockin cache */

char *currpos; /* currentinsertion point in cache */

XLogRecPtr RedoRecPtr; /*current redo point for insertions */

bool forcePageWrites; /* forcing full-page writes for PITR? */

/*

* exclusiveBackup is true if a backup startedwith pg_start_backup() is

* in progress, and nonExclusiveBackups is acounter indicating the number

* of streaming base backups currently inprogress. forcePageWrites is set

* to true when either of these is non-zero.lastBackupStart is the latest

* checkpoint redo location used as a startingpoint for an online backup.

*/

bool exclusiveBackup;

int nonExclusiveBackups;

XLogRecPtr lastBackupStart;

} XLogCtlInsert;

在XLOGSHmemInit()函数里，首先在shmem的哈希表索引"ShmemIndex"上给控制文件pg_control增加一个HashElement和ShmemIndexEnt（entry），在shmem里根据ControlFileData大小调用ShmemAlloc()分配内存空间，使ShmemIndexEnt的成员location指向该空间，size成员记录该空间大小。

XLOGSHmemInit()调用ShmemInitStruct()，在其中调用hash_search()在哈希表索引"ShmemIndex"中查找"XLOGCtl"，如果没有，就在shmemIndex中给"XLOG Ctl"分一个HashElement和ShmemIndexEnt（entry），在其中的Entry中写上"XLOG Ctl"。返回ShmemInitStruct()，再调用ShmemAlloc()在共享内存上给"XLOG Ctl"相关结构（见下面“XLog相关结构图”）分配空间，设置entry（在这儿及ShmemIndexEnt类型变量）的成员location指向该空间，size成员记录该空间大小，最后返回XLOGShmemInit()，让XLogCtlData *类型静态全局变量XLogCtl指向在shmem里给"XLOG Ctl"相关结构分配的内存地址，设置其中XLogCtlData结构类型的成员值。初始化完成后数据结构如下图。

初始化完xlog的内存结构图

为了精简上图，把创建shmem的哈希表索引"ShmemIndex"时创建的HCTL结构删掉了，这个结构的作用是记录创建可扩展哈希表的相关信息。增加了左边灰色底的部分，描述共享内存/shmem里各变量物理布局概览，由下往上，由低地址到高地址。其中的"Control File"即ControlFileDate和"XLOG Ctl"即xlog的相关结构图下面分别给出，要不上面的图太大了。

控制文件结构图

上图中ControlFileData结构中的XLogRecPtr和CheckPoint不是指针，因此应该用右边的相应结构图代替，把这两个合进去有点费劲，将就着看吧。

XLog相关结构图