PostgreSQL 原始碼解讀（105）- WAL#2（Insert & WAL-heap_i...

husthxd發表於2018-12-10
SQL原始碼

本節介紹了插入資料時與WAL相關的處理邏輯，主要包括heap_insert依賴的函式XLogBeginInsert/XLogRegisterBufData/XLogRegisterData/XLogSetRecordFlags。

一、資料結構

靜態變數
程式中全域性共享

/*
 * An array of XLogRecData structs, to hold registered data.
 * XLogRecData結構體陣列,儲存已註冊的資料
 */
static XLogRecData *rdatas;
//已使用的入口
static int  num_rdatas;         /* entries currently used */
//已分配的空間大小
static int  max_rdatas;         /* allocated size */
//是否呼叫XLogBeginInsert函式
static bool begininsert_called = false;

registered_buffer
對於每一個使用XLogRegisterBuffer註冊的每個資料塊,填充到registered_buffer結構體中


/*
 * For each block reference registered with XLogRegisterBuffer, we fill in
 * a registered_buffer struct.
 * 對於每一個使用XLogRegisterBuffer註冊的每個資料塊,
 *   填充到registered_buffer結構體中
 */
typedef struct
{
    //slot是否在使用?
    bool        in_use;         /* is this slot in use? */
    //REGBUF_* 相關標記
    uint8       flags;          /* REGBUF_* flags */
    //定義關係和資料庫的識別符號
    RelFileNode rnode;          /* identifies the relation and block */
    //fork程式編號
    ForkNumber  forkno;
    //塊編號
    BlockNumber block;
    //頁內容
    Page        page;           /* page content */
    //rdata鏈中的資料總大小
    uint32      rdata_len;      /* total length of data in rdata chain */
    //使用該資料塊註冊的資料鏈頭
    XLogRecData *rdata_head;    /* head of the chain of data registered with
                                 * this block */
    //使用該資料塊註冊的資料鏈尾
    XLogRecData *rdata_tail;    /* last entry in the chain, or &rdata_head if
                                 * empty */
    //臨時rdatas資料引用,用於儲存XLogRecordAssemble()中使用的備份塊資料
    XLogRecData bkp_rdatas[2];  /* temporary rdatas used to hold references to
                                 * backup block data in XLogRecordAssemble() */

    /* buffer to store a compressed version of backup block image */
    //用於儲存壓縮版本的備份塊映象的快取
    char        compressed_page[PGLZ_MAX_BLCKSZ];
} registered_buffer;
//registered_buffer指正
static registered_buffer *registered_buffers;
//已分配的大小
static int  max_registered_buffers; /* allocated size */
//最大塊號 + 1(當前註冊塊)
static int  max_registered_block_id = 0;    /* highest block_id + 1 currently
                                             * registered */

XLogCtlInsert
WAL插入記錄時使用的共享資料結構

/*
 * Shared state data for WAL insertion.
 * WAL插入記錄時使用的共享資料結構
 */
typedef struct XLogCtlInsert
{
    //包含CurrBytePos和PrevBytePos的lock
    slock_t     insertpos_lck;  /* protects CurrBytePos and PrevBytePos */

    /*
     * CurrBytePos is the end of reserved WAL. The next record will be
     * inserted at that position. PrevBytePos is the start position of the
     * previously inserted (or rather, reserved) record - it is copied to the
     * prev-link of the next record. These are stored as "usable byte
     * positions" rather than XLogRecPtrs (see XLogBytePosToRecPtr()).
     * CurrBytePos是保留WAL的結束位置。
     *   下一條記錄將插入到那個位置。
     * PrevBytePos是先前插入(或者保留)記錄的起始位置——它被複制到下一條記錄的prev-link中。
     * 這些儲存為“可用位元組位置”，而不是XLogRecPtrs(參見XLogBytePosToRecPtr())。
     */
    uint64      CurrBytePos;
    uint64      PrevBytePos;

    /*
     * Make sure the above heavily-contended spinlock and byte positions are
     * on their own cache line. In particular, the RedoRecPtr and full page
     * write variables below should be on a different cache line. They are
     * read on every WAL insertion, but updated rarely, and we don't want
     * those reads to steal the cache line containing Curr/PrevBytePos.
     * 確保以上激烈競爭的自旋鎖和位元組位置在它們自己的快取line上。
     * 特別是，RedoRecPtr和下面的全頁寫變數應該位於不同的快取line上。
     * 它們在每次插入WAL時都被讀取，但很少更新，
     *   我們不希望這些讀取竊取包含Curr/PrevBytePos的快取line。
     */
    char        pad[PG_CACHE_LINE_SIZE];

    /*
     * fullPageWrites is the master copy used by all backends to determine
     * whether to write full-page to WAL, instead of using process-local one.
     * This is required because, when full_page_writes is changed by SIGHUP,
     * we must WAL-log it before it actually affects WAL-logging by backends.
     * Checkpointer sets at startup or after SIGHUP.
     * fullpagewrite是所有後臺程式使用的主副本，
     *   用於確定是否將整個頁面寫入WAL，而不是使用process-local副本。
     * 這是必需的，因為當SIGHUP更改full_page_write時，
     *   我們必須在它透過後臺程式實際影響WAL-logging之前對其進行WAL-log記錄。
     * Checkpointer檢查點設定在啟動或SIGHUP之後。
     *
     * To read these fields, you must hold an insertion lock. To modify them,
     * you must hold ALL the locks.
     * 為了讀取這些域,必須持有insertion lock.
     * 如需更新,則需要持有所有這些lock. 
     */
    //插入時的當前redo point
    XLogRecPtr  RedoRecPtr;     /* current redo point for insertions */
    //為PITR強制執行full-page寫?
    bool        forcePageWrites;    /* forcing full-page writes for PITR? */
    //是否全頁寫?
    bool        fullPageWrites;

    /*
     * exclusiveBackupState indicates the state of an exclusive backup (see
     * comments of ExclusiveBackupState for more details). nonExclusiveBackups
     * is a counter indicating the number of streaming base backups currently
     * in progress. forcePageWrites is set to true when either of these is
     * non-zero. lastBackupStart is the latest checkpoint redo location used
     * as a starting point for an online backup.
     * exclusive sivebackupstate表示排他備份的狀態
     * (有關詳細資訊，請參閱exclusive sivebackupstate的註釋)。
     * 非排他性備份是一個計數器，指示當前正在進行的流基礎備份的數量。
     * forcePageWrites在這兩個值都不為零時被設定為true。
     * lastBackupStart用作線上備份起點的最新檢查點的重做位置。
     */
    ExclusiveBackupState exclusiveBackupState;
    int         nonExclusiveBackups;
    XLogRecPtr  lastBackupStart;

    /*
     * WAL insertion locks.
     * WAL寫入鎖
     */
    WALInsertLockPadded *WALInsertLocks;
} XLogCtlInsert;

XLogRecData
xloginsert.c中的函式構造一個XLogRecData結構體鏈用於標識最後的WAL記錄

/*
 * The functions in xloginsert.c construct a chain of XLogRecData structs
 * to represent the final WAL record.
 * xloginsert.c中的函式構造一個XLogRecData結構體鏈用於標識最後的WAL記錄
 */
typedef struct XLogRecData
{
    //鏈中的下一個結構體,如無則為NULL
    struct XLogRecData *next;   /* next struct in chain, or NULL */
    //rmgr資料的起始地址
    char       *data;           /* start of rmgr data to include */
    //rmgr資料大小
    uint32      len;            /* length of rmgr data to include */
} XLogRecData;

registered_buffer/registered_buffers
對於每一個使用XLogRegisterBuffer註冊的每個資料塊,填充到registered_buffer結構體中


/*
 * For each block reference registered with XLogRegisterBuffer, we fill in
 * a registered_buffer struct.
 * 對於每一個使用XLogRegisterBuffer註冊的每個資料塊,
 *   填充到registered_buffer結構體中
 */
typedef struct
{
    //slot是否在使用?
    bool        in_use;         /* is this slot in use? */
    //REGBUF_* 相關標記
    uint8       flags;          /* REGBUF_* flags */
    //定義關係和資料庫的識別符號
    RelFileNode rnode;          /* identifies the relation and block */
    //fork程式編號
    ForkNumber  forkno;
    //塊編號
    BlockNumber block;
    //頁內容
    Page        page;           /* page content */
    //rdata鏈中的資料總大小
    uint32      rdata_len;      /* total length of data in rdata chain */
    //使用該資料塊註冊的資料鏈頭
    XLogRecData *rdata_head;    /* head of the chain of data registered with
                                 * this block */
    //使用該資料塊註冊的資料鏈尾
    XLogRecData *rdata_tail;    /* last entry in the chain, or &rdata_head if
                                 * empty */
    //臨時rdatas資料引用,用於儲存XLogRecordAssemble()中使用的備份塊資料
    XLogRecData bkp_rdatas[2];  /* temporary rdatas used to hold references to
                                 * backup block data in XLogRecordAssemble() */

    /* buffer to store a compressed version of backup block image */
    //用於儲存壓縮版本的備份塊映象的快取
    char        compressed_page[PGLZ_MAX_BLCKSZ];
} registered_buffer;
//registered_buffer指標(全域性變數)
static registered_buffer *registered_buffers;
//已分配的大小
static int  max_registered_buffers; /* allocated size */
//最大塊號 + 1(當前註冊塊)
static int  max_registered_block_id = 0;    /* highest block_id + 1 currently
                                             * registered */

二、原始碼解讀

heap_insert
主要實現邏輯是插入元組到堆中,其中存在對WAL(XLog)進行處理的部分.
參見PostgreSQL 原始碼解讀（104）- WAL#1（Insert & WAL-heap_insert函式#1）

XLogBeginInsert
開始構造WAL記錄.
必須在呼叫XLogRegister*和XLogInsert()函式前呼叫.

/*
 * Begin constructing a WAL record. This must be called before the
 * XLogRegister* functions and XLogInsert().
 * 開始構造WAL記錄.
 * 必須在呼叫XLogRegister*和XLogInsert()函式前呼叫.
 */
void
XLogBeginInsert(void)
{
    //驗證邏輯
    Assert(max_registered_block_id == 0);
    Assert(mainrdata_last == (XLogRecData *) &mainrdata_head);
    Assert(mainrdata_len == 0);

    /* cross-check on whether we should be here or not */
    //交叉校驗是否應該在這裡還是不應該在這裡出現
    if (!XLogInsertAllowed())
        elog(ERROR, "cannot make new WAL entries during recovery");

    if (begininsert_called)
        elog(ERROR, "XLogBeginInsert was already called");
    //變數賦值
    begininsert_called = true;
}

/*
 * Is this process allowed to insert new WAL records?
 * 判斷該程式是否允許插入新的WAL記錄
 * 
 * Ordinarily this is essentially equivalent to !RecoveryInProgress().
 * But we also have provisions for forcing the result "true" or "false"
 * within specific processes regardless of the global state.
 * 通常，這本質上等同於! recoverinprogress()。
 * 但我們也有規定，無論全域性狀況如何，都要在特定程式中強制實現“正確”或“錯誤”的結果。
 */
bool
XLogInsertAllowed(void)
{
    /*
     * If value is "unconditionally true" or "unconditionally false", just
     * return it.  This provides the normal fast path once recovery is known
     * done.
     * 如果值為“無條件為真”或“無條件為假”，則返回。
     * 這提供正常的快速判斷路徑。
     */
    if (LocalXLogInsertAllowed >= 0)
        return (bool) LocalXLogInsertAllowed;

    /*
     * Else, must check to see if we're still in recovery.
     * 否則,必須檢查是否處於恢復狀態
     */
    if (RecoveryInProgress())
        return false;

    /*
     * On exit from recovery, reset to "unconditionally true", since there is
     * no need to keep checking.
     * 從恢復中退出,由於不需要繼續檢查,重置為"無條件為真"
     */
    LocalXLogInsertAllowed = 1;
    return true;
}

XLogRegisterData
新增資料到正在構造的WAL記錄中

/*
 * Add data to the WAL record that's being constructed.
 * 新增資料到正在構造的WAL記錄中
 * 
 * The data is appended to the "main chunk", available at replay with
 * XLogRecGetData().
 * 資料追加到"main chunk"中,用於XLogRecGetData()函式回放
 */
void
XLogRegisterData(char *data, int len)
{
    XLogRecData *rdata;//資料
    //驗證是否已呼叫begin
    Assert(begininsert_called);
    //驗證大小
    if (num_rdatas >= max_rdatas)
        elog(ERROR, "too much WAL data");
    rdata = &rdatas[num_rdatas++];

    rdata->data = data;
    rdata->len = len;

    /*
     * we use the mainrdata_last pointer to track the end of the chain, so no
     * need to clear 'next' here.
     * 使用mainrdata_last指標跟蹤鏈條的結束點,在這裡不需要清除next變數
     */

    mainrdata_last->next = rdata;
    mainrdata_last = rdata;

    mainrdata_len += len;
}

XLogRegisterBuffer
在緩衝區中註冊已構建的WAL記錄的依賴,在WAL-logged操作更新每一個page時必須呼叫此函式

/*
 * Register a reference to a buffer with the WAL record being constructed.
 * This must be called for every page that the WAL-logged operation modifies.
 * 在緩衝區中註冊已構建的WAL記錄的依賴
 * 在WAL-logged操作更新每一個page時必須呼叫此函式
 */
void
XLogRegisterBuffer(uint8 block_id, Buffer buffer, uint8 flags)
{
    registered_buffer *regbuf;//緩衝

    /* NO_IMAGE doesn't make sense with FORCE_IMAGE */
    //NO_IMAGE不能與REGBUF_NO_IMAGE同時使用
    Assert(!((flags & REGBUF_FORCE_IMAGE) && (flags & (REGBUF_NO_IMAGE))));
    Assert(begininsert_called);
    //塊ID > 最大已註冊的緩衝區,報錯
    if (block_id >= max_registered_block_id)
    {
        if (block_id >= max_registered_buffers)
            elog(ERROR, "too many registered buffers");
        max_registered_block_id = block_id + 1;
    }
    //賦值
    regbuf = &registered_buffers[block_id];
    //獲取Tag
    BufferGetTag(buffer, &regbuf->rnode, &regbuf->forkno, &regbuf->block);
    regbuf->page = BufferGetPage(buffer);
    regbuf->flags = flags;
    regbuf->rdata_tail = (XLogRecData *) &regbuf->rdata_head;
    regbuf->rdata_len = 0;

    /*
     * Check that this page hasn't already been registered with some other
     * block_id.
     * 檢查該page是否已被其他block_id註冊
     */
#ifdef USE_ASSERT_CHECKING
    {
        int         i;

        for (i = 0; i < max_registered_block_id; i++)//迴圈檢查
        {
            registered_buffer *regbuf_old = &registered_buffers[i];

            if (i == block_id || !regbuf_old->in_use)
                continue;

            Assert(!RelFileNodeEquals(regbuf_old->rnode, regbuf->rnode) ||
                   regbuf_old->forkno != regbuf->forkno ||
                   regbuf_old->block != regbuf->block);
        }
    }
#endif

    regbuf->in_use = true;//標記為使用
}


/*
 * BufferGetTag
 *      Returns the relfilenode, fork number and block number associated with
 *      a buffer.
 * 返回與緩衝區相關的relfilenode,fork編號和塊號
 */
void
BufferGetTag(Buffer buffer, RelFileNode *rnode, ForkNumber *forknum,
             BlockNumber *blknum)
{
    BufferDesc *bufHdr;

    /* Do the same checks as BufferGetBlockNumber. */
    //驗證buffer已被pinned
    Assert(BufferIsPinned(buffer));

    if (BufferIsLocal(buffer))
        bufHdr = GetLocalBufferDescriptor(-buffer - 1);
    else
        bufHdr = GetBufferDescriptor(buffer - 1);

    /* pinned, so OK to read tag without spinlock */
    //pinned,不需要spinlock讀取tage
    *rnode = bufHdr->tag.rnode;
    *forknum = bufHdr->tag.forkNum;
    *blknum = bufHdr->tag.blockNum;
}

/*
 * BufferIsLocal
 *      True iff the buffer is local (not visible to other backends).
 *      如緩衝區對其他後臺程式不不可見,則為本地buffer
 */
#define BufferIsLocal(buffer)   ((buffer) < 0)
#define GetBufferDescriptor(id) (&BufferDescriptors[(id)].bufferdesc)
#define GetLocalBufferDescriptor(id) (&LocalBufferDescriptors[(id)])
BufferDesc *LocalBufferDescriptors = NULL;
BufferDescPadded *BufferDescriptors;

XLogRegisterBufData
在正在構造的WAL記錄中新增buffer相關的資料.

/*
 * Add buffer-specific data to the WAL record that's being constructed.
 * 在正在構造的WAL記錄中新增buffer相關的資料.
 * 
 * Block_id must reference a block previously registered with
 * XLogRegisterBuffer(). If this is called more than once for the same
 * block_id, the data is appended.
 * Block_id必須引用先前註冊到XLogRegisterBuffer()中的資料塊。
 * 如果對同一個block_id不止一次呼叫，那麼資料將會追加。
 *
 * The maximum amount of data that can be registered per block is 65535
 * bytes. That should be plenty; if you need more than BLCKSZ bytes to
 * reconstruct the changes to the page, you might as well just log a full
 * copy of it. (the "main data" that's not associated with a block is not
 * limited)
 * 每個塊可註冊的最大大小是65535Bytes.
 * 通常來說這已經足夠了;如果需要大小比BLCKSZ位元組更大的資料用於重建頁面的變化,
 *   那麼需要整頁進行複製.
 * (與資料塊相關的"main data"是不受限的)
 */
void
XLogRegisterBufData(uint8 block_id, char *data, int len)
{
    registered_buffer *regbuf;//註冊的緩衝區
    XLogRecData *rdata;//資料

    Assert(begininsert_called);//XLogBeginInsert函式已呼叫

    /* find the registered buffer struct */
    //尋找已註冊的快取結構體
    regbuf = &registered_buffers[block_id];
    if (!regbuf->in_use)
        elog(ERROR, "no block with id %d registered with WAL insertion",
             block_id);

    if (num_rdatas >= max_rdatas)
        elog(ERROR, "too much WAL data");
    rdata = &rdatas[num_rdatas++];

    rdata->data = data;
    rdata->len = len;

    regbuf->rdata_tail->next = rdata;
    regbuf->rdata_tail = rdata;
    regbuf->rdata_len += len;
}

XLogSetRecordFlags
為即將"到來"的WAL記錄設定插入狀態標記
XLOG_INCLUDE_ORIGIN 確定複製起點是否應該包含在記錄中
XLOG_MARK_UNIMPORTANT 表示記錄對於永續性並不重要，這可以避免觸發WAL歸檔和其他後臺活動

/*
 * Set insert status flags for the upcoming WAL record.
 * 為即將"到來"的WAL記錄設定插入狀態標記
 *
 * The flags that can be used here are:
 * - XLOG_INCLUDE_ORIGIN, to determine if the replication origin should be
 *   included in the record.
 * - XLOG_MARK_UNIMPORTANT, to signal that the record is not important for
 *   durability, which allows to avoid triggering WAL archiving and other
 *   background activity.
 * 標記用於:
 * - XLOG_INCLUDE_ORIGIN 確定複製起點是否應該包含在記錄中
 * - XLOG_MARK_UNIMPORTANT 表示記錄對於永續性並不重要，這可以避免觸發WAL歸檔和其他後臺活動。
 */
void
XLogSetRecordFlags(uint8 flags)
{
    Assert(begininsert_called);
    curinsert_flags = flags;
}

三、跟蹤分析

測試指令碼如下

insert into t_wal_partition(c1,c2,c3) VALUES(0,'HASH0','HAHS0');

XLogBeginInsert
啟動gdb,設定斷點,進入XLogBeginInsert

(gdb) b XLogBeginInsert
Breakpoint 1 at 0x564897: file xloginsert.c, line 122.
(gdb) c
Continuing.

Breakpoint 1, XLogBeginInsert () at xloginsert.c:122
122     Assert(max_registered_block_id == 0);

校驗,呼叫XLogInsertAllowed

122     Assert(max_registered_block_id == 0);
(gdb) n
123     Assert(mainrdata_last == (XLogRecData *) &mainrdata_head);
(gdb) 
124     Assert(mainrdata_len == 0);
(gdb) 
127     if (!XLogInsertAllowed())
(gdb) step
XLogInsertAllowed () at xlog.c:8126
8126        if (LocalXLogInsertAllowed >= 0)
(gdb) n
8132        if (RecoveryInProgress())
(gdb) 
8139        LocalXLogInsertAllowed = 1;
(gdb) 
8140        return true;
(gdb) 
8141    }
(gdb)

賦值,設定begininsert_called為T,返回

(gdb) 
XLogBeginInsert () at xloginsert.c:130
130     if (begininsert_called)
(gdb) p begininsert_called
$1 = false
(gdb) n
133     begininsert_called = true;
(gdb) 
134 }
(gdb) 
heap_insert (relation=0x7f5cc0338228, tup=0x29b2440, cid=0, options=0, bistate=0x0) at heapam.c:2567
2567            XLogRegisterData((char *) &xlrec, SizeOfHeapInsert);
(gdb)

XLogRegisterData
進入XLogRegisterData函式

(gdb) step
XLogRegisterData (data=0x7fff03ba99e0 "\002", len=3) at xloginsert.c:327
327     Assert(begininsert_called);
(gdb) p *data
$2 = 2 '\002'
(gdb) p *(xl_heap_insert *)data
$3 = {offnum = 2, flags = 0 '\000'}

執行相關判斷,並賦值
rdatas是XLogRecData結構體指標,全域性靜態變數:
static XLogRecData *rdatas;

(gdb) n
329     if (num_rdatas >= max_rdatas)
(gdb) p num_rdatas
$4 = 0
(gdb) p max_rdatas
$5 = 20
(gdb) n
331     rdata = &rdatas[num_rdatas++];
(gdb) p rdatas[0]
$6 = {next = 0x0, data = 0x0, len = 0}
(gdb) p rdatas[1]
$7 = {next = 0x0, data = 0x0, len = 0}

相關結構體賦值
其中mainrdata_last是mainrdata_head的地址:
static XLogRecData *mainrdata_head;
static XLogRecData *mainrdata_last = (XLogRecData *) &mainrdata_head;

(gdb) n
333     rdata->data = data;
(gdb) 
334     rdata->len = len;
(gdb) 
341     mainrdata_last->next = rdata;
(gdb) 
342     mainrdata_last = rdata;
(gdb) 
344     mainrdata_len += len;
(gdb) 
345 }

完成呼叫,回到heap_insert

(gdb) n
heap_insert (relation=0x7f5cc0338228, tup=0x29b2440, cid=0, options=0, bistate=0x0) at heapam.c:2569
2569            xlhdr.t_infomask2 = heaptup->t_data->t_infomask2;

XLogRegisterBuffer
進入XLogRegisterBuffer

(gdb) step
XLogRegisterBuffer (block_id=0 '\000', buffer=99, flags=8 '\b') at xloginsert.c:218
218     Assert(!((flags & REGBUF_FORCE_IMAGE) && (flags & (REGBUF_NO_IMAGE))));

判斷block_id,設定max_registered_block_id變數等.
注:max_registered_buffers初始化為5

(gdb) n
219     Assert(begininsert_called);
(gdb) 
221     if (block_id >= max_registered_block_id)
(gdb) p max_registered_block_id
$14 = 0
(gdb) n
223         if (block_id >= max_registered_buffers)
(gdb) p max_registered_buffers
$15 = 5
(gdb) n
225         max_registered_block_id = block_id + 1;
(gdb) 
228     regbuf = &registered_buffers[block_id];
(gdb) p max_registered_buffers
$16 = 5
(gdb) p max_registered_block_id
$17 = 1
(gdb) n
230     BufferGetTag(buffer, &regbuf->rnode, &regbuf->forkno, &regbuf->block);
(gdb) p *regbuf
$18 = {in_use = false, flags = 0 '\000', rnode = {spcNode = 0, dbNode = 0, relNode = 0}, forkno = MAIN_FORKNUM, block = 0, 
  page = 0x0, rdata_len = 0, rdata_head = 0x0, rdata_tail = 0x0, bkp_rdatas = {{next = 0x0, data = 0x0, len = 0}, {
      next = 0x0, data = 0x0, len = 0}}, compressed_page = '\000' <repeats 8195 times>}

獲取buffer的tag
rnode/forkno/block

(gdb) n
231     regbuf->page = BufferGetPage(buffer);
(gdb) p *regbuf
$19 = {in_use = false, flags = 0 '\000', rnode = {spcNode = 1663, dbNode = 16402, relNode = 17034}, forkno = MAIN_FORKNUM, 
  block = 0, page = 0x0, rdata_len = 0, rdata_head = 0x0, rdata_tail = 0x0, bkp_rdatas = {{next = 0x0, data = 0x0, 
      len = 0}, {next = 0x0, data = 0x0, len = 0}}, compressed_page = '\000' <repeats 8195 times>}

設定flags等其他變數

(gdb) n
232     regbuf->flags = flags;
(gdb) 
233     regbuf->rdata_tail = (XLogRecData *) &regbuf->rdata_head;
(gdb) 
234     regbuf->rdata_len = 0;
(gdb) 
244         for (i = 0; i < max_registered_block_id; i++)
(gdb) p regbuf->flags
$21 = 8 '\b'
(gdb) p *regbuf->rdata_tail
$23 = {next = 0x0, data = 0x292e1a8 "", len = 0}
(gdb) p regbuf->rdata_len
$24 = 0

檢查該page是否已被其他block_id註冊
最後設定in_use為T,返回XLogRegisterBufData

(gdb) n
246             registered_buffer *regbuf_old = &registered_buffers[i];
(gdb) 
248             if (i == block_id || !regbuf_old->in_use)
(gdb) 
249                 continue;
(gdb) 
244         for (i = 0; i < max_registered_block_id; i++)
(gdb) 
258     regbuf->in_use = true;
(gdb) 
259 }
(gdb) 
heap_insert (relation=0x7f5cc0338228, tup=0x29b2440, cid=0, options=0, bistate=0x0) at heapam.c:2579
2579            XLogRegisterBufData(0, (char *) &xlhdr, SizeOfHeapHeader);

XLogRegisterBufData
進入XLogRegisterBufData函式

(gdb) step
XLogRegisterBufData (block_id=0 '\000', data=0x7fff03ba99d0 "\003", len=5) at xloginsert.c:366
366     Assert(begininsert_called);

尋找已註冊的快取結構體

(gdb) n
369     regbuf = &registered_buffers[block_id];
(gdb) 
370     if (!regbuf->in_use)
(gdb) p *regbuf
$25 = {in_use = true, flags = 8 '\b', rnode = {spcNode = 1663, dbNode = 16402, relNode = 17034}, forkno = MAIN_FORKNUM, 
  block = 0, page = 0x7f5c93854380 "\001", rdata_len = 0, rdata_head = 0x0, rdata_tail = 0x292e1a8, bkp_rdatas = {{
      next = 0x0, data = 0x0, len = 0}, {next = 0x0, data = 0x0, len = 0}}, compressed_page = '\000' <repeats 8195 times>}
(gdb) p *regbuf->page
$26 = 1 '\001'
(gdb) n
374     if (num_rdatas >= max_rdatas)
(gdb)

在正在構造的WAL記錄中新增buffer相關的資料.

(gdb) n
376     rdata = &rdatas[num_rdatas++];
(gdb) p num_rdatas
$27 = 1
(gdb) p max_rdatas
$28 = 20
(gdb) n
378     rdata->data = data;
(gdb) 
379     rdata->len = len;
(gdb) 
381     regbuf->rdata_tail->next = rdata;
(gdb) 
382     regbuf->rdata_tail = rdata;
(gdb) 
383     regbuf->rdata_len += len;
(gdb) 
384 }
(gdb) p *rdata
$29 = {next = 0x0, data = 0x7fff03ba99d0 "\003", len = 5}
(gdb)

完成呼叫,回到heap_insert

(gdb) n
heap_insert (relation=0x7f5cc0338228, tup=0x29b2440, cid=0, options=0, bistate=0x0) at heapam.c:2583
2583                                heaptup->t_len - SizeofHeapTupleHeader);

繼續呼叫XLogRegisterBufData函式註冊tuple實際資料

2583                                heaptup->t_len - SizeofHeapTupleHeader);
(gdb) n
2581            XLogRegisterBufData(0,
(gdb)

XLogSetRecordFlags
為即將"到來"的WAL記錄設定插入狀態標記

(gdb) 
2586            XLogSetRecordFlags(XLOG_INCLUDE_ORIGIN);

邏輯很簡單,設定標記位curinsert_flags

(gdb) step
XLogSetRecordFlags (flags=1 '\001') at xloginsert.c:399
399     Assert(begininsert_called);
(gdb) n
400     curinsert_flags = flags;
(gdb) 
401 }
(gdb) 
heap_insert (relation=0x7f5cc0338228, tup=0x29b2440, cid=0, options=0, bistate=0x0) at heapam.c:2588
2588            recptr = XLogInsert(RM_HEAP_ID, info);
(gdb)

呼叫XLogInsert,插入WAL

(gdb) 
2590            PageSetLSN(page, recptr);
...

XLogInsert函式下節再行介紹.

四、參考資料

Write Ahead Logging — WAL
PostgreSQL 原始碼解讀（4）- 插入資料#3（heap_insert）
PgSQL · 特性分析 · 資料庫崩潰恢復（上）
PgSQL · 特性分析 · 資料庫崩潰恢復（下）
PgSQL · 特性分析 · Write-Ahead Logging機制淺析
PostgreSQL WAL Buffers, Clog Buffers Deep Dive

來自 “ ITPUB部落格 ” ，連結：http://blog.itpub.net/6906/viewspace-2374784/，如需轉載，請註明出處，否則將追究法律責任。

相關文章