PostgreSQL 源码解读（113）- WAL#9（Inse
2018-12-29 本文已影响13人
EthanHe
本节重点跟踪分析了ReserveXLogInsertLocation和CopyXLogRecordToWAL函数的实现逻辑,ReserveXLogInsertLocation函数为XLOG Record预留合适的空间,CopyXLogRecordToWAL则负责拷贝XLOG Record到WAL buffer的保留空间中。
一、数据结构
全局变量
/* flags for the in-progress insertion */
//用于插入过程中的标记信息
static uint8 curinsert_flags = 0;
/*
* These are used to hold the record header while constructing a record.
* 'hdr_scratch' is not a plain variable, but is palloc'd at initialization,
* because we want it to be MAXALIGNed and padding bytes zeroed.
* 在构建XLOG Record时通常会存储记录的头部信息.
* 'hdr_scratch'并不是一个普通(plain)变量,而是在初始化时通过palloc初始化,
* 因为我们希望该变量已经是MAXALIGNed并且已被0x00填充.
*
* For simplicity, it's allocated large enough to hold the headers for any
* WAL record.
* 简单起见,该变量预先会分配足够大的空间用于存储所有WAL Record的头部信息.
*/
static XLogRecData hdr_rdt;
static char *hdr_scratch = NULL;
#define SizeOfXlogOrigin (sizeof(RepOriginId) + sizeof(char))
#define HEADER_SCRATCH_SIZE \
(SizeOfXLogRecord + \
MaxSizeOfXLogRecordBlockHeader * (XLR_MAX_BLOCK_ID + 1) + \
SizeOfXLogRecordDataHeaderLong + SizeOfXlogOrigin)
/*
* An array of XLogRecData structs, to hold registered data.
* XLogRecData结构体数组,存储已注册的数据.
*/
static XLogRecData *rdatas;
static int num_rdatas; /* entries currently used */
//已分配的空间大小
static int max_rdatas; /* allocated size */
//是否调用XLogBeginInsert函数
static bool begininsert_called = false;
static XLogCtlData *XLogCtl = NULL;
/* flags for the in-progress insertion */
static uint8 curinsert_flags = 0;
/*
* A chain of XLogRecDatas to hold the "main data" of a WAL record, registered
* with XLogRegisterData(...).
* 存储WAL Record "main data"的XLogRecDatas数据链
*/
static XLogRecData *mainrdata_head;
static XLogRecData *mainrdata_last = (XLogRecData *) &mainrdata_head;
//链中某个位置的mainrdata大小
static uint32 mainrdata_len; /* total # of bytes in chain */
/*
* ProcLastRecPtr points to the start of the last XLOG record inserted by the
* current backend. It is updated for all inserts. XactLastRecEnd points to
* end+1 of the last record, and is reset when we end a top-level transaction,
* or start a new one; so it can be used to tell if the current transaction has
* created any XLOG records.
* ProcLastRecPtr指向当前后端插入的最后一条XLOG记录的开头。
* 它针对所有插入进行更新。
* XactLastRecEnd指向最后一条记录的末尾位置 + 1，
* 并在结束顶级事务或启动新事务时重置;
* 因此，它可以用来判断当前事务是否创建了任何XLOG记录。
*
* While in parallel mode, this may not be fully up to date. When committing,
* a transaction can assume this covers all xlog records written either by the
* user backend or by any parallel worker which was present at any point during
* the transaction. But when aborting, or when still in parallel mode, other
* parallel backends may have written WAL records at later LSNs than the value
* stored here. The parallel leader advances its own copy, when necessary,
* in WaitForParallelWorkersToFinish.
* 在并行模式下，这可能不是完全是最新的。
* 在提交时，事务可以假定覆盖了用户后台进程或在事务期间出现的并行worker进程的所有xlog记录。
* 但是，当中止时，或者仍然处于并行模式时，其他并行后台进程可能在较晚的LSNs中写入了WAL记录，
* 而不是存储在这里的值。
* 当需要时，并行处理进程的leader在WaitForParallelWorkersToFinish中会推进自己的副本。
*/
XLogRecPtr ProcLastRecPtr = InvalidXLogRecPtr;
XLogRecPtr XactLastRecEnd = InvalidXLogRecPtr;
XLogRecPtr XactLastCommitEnd = InvalidXLogRecPtr;
/* For WALInsertLockAcquire/Release functions */
//用于WALInsertLockAcquire/Release函数
static int MyLockNo = 0;
static bool holdingAllLocks = false;
/*
* Private, possibly out-of-date copy of shared LogwrtResult.
* See discussion above.
* 进程私有的可能已过期的共享LogwrtResult变量的拷贝.
*/
static XLogwrtResult LogwrtResult = {0, 0};
/* The number of bytes in a WAL segment usable for WAL data. */
//WAL segment file中可用于WAL data的字节数(不包括page header)
static int UsableBytesInSegment;
宏定义
XLogRegisterBuffer函数使用的flags
/* flags for XLogRegisterBuffer */
//XLogRegisterBuffer函数使用的flags
#define REGBUF_FORCE_IMAGE 0x01 /* 强制执行full-page-write;force a full-page image */
#define REGBUF_NO_IMAGE 0x02 /* 不需要FPI;don't take a full-page image */
#define REGBUF_WILL_INIT (0x04 | 0x02) /* 在回放时重新初始化page(表示NO_IMAGE);
* page will be re-initialized at
* replay (implies NO_IMAGE) */
#define REGBUF_STANDARD 0x08 /* 标准的page layout(数据在pd_lower和pd_upper之间的数据会被跳过)
* page follows "standard" page layout,
* (data between pd_lower and pd_upper
* will be skipped) */
#define REGBUF_KEEP_DATA 0x10 /* include data even if a full-page image
* is taken */
/*
* Flag bits for the record being inserted, set using XLogSetRecordFlags().
*/
#define XLOG_INCLUDE_ORIGIN 0x01 /* include the replication origin */
#define XLOG_MARK_UNIMPORTANT 0x02 /* record not important for durability */
#define XLogSegmentOffset(xlogptr, wal_segsz_bytes) \
((xlogptr) & ((wal_segsz_bytes) - 1))
/*
* Calculate the amount of space left on the page after 'endptr'. Beware
* multiple evaluation!
* 计算page中在"endptr"后的剩余空闲空间.注意multiple evaluation!
*/
#define INSERT_FREESPACE(endptr) \
(((endptr) % XLOG_BLCKSZ == 0) ? 0 : (XLOG_BLCKSZ - (endptr) % XLOG_BLCKSZ))
XLogRecData
xloginsert.c中的函数构造一个XLogRecData结构体链用于标识最后的WAL记录
/*
* The functions in xloginsert.c construct a chain of XLogRecData structs
* to represent the final WAL record.
* xloginsert.c中的函数构造一个XLogRecData结构体链用于标识最后的WAL记录
*/
typedef struct XLogRecData
{
//链中的下一个结构体,如无则为NULL
struct XLogRecData *next; /* next struct in chain, or NULL */
//rmgr数据的起始地址
char *data; /* start of rmgr data to include */
//rmgr数据大小
uint32 len; /* length of rmgr data to include */
} XLogRecData;
二、源码解读
ReserveXLogInsertLocation
在WAL(buffer)中为给定大小的记录预留合适的空间。*StartPos设置为预留部分的开头，*EndPos设置为其结尾+1。*PrePtr设置为前一记录的开头;它用于设置该记录的xl_prev变量。
/*
* Reserves the right amount of space for a record of given size from the WAL.
* *StartPos is set to the beginning of the reserved section, *EndPos to
* its end+1. *PrevPtr is set to the beginning of the previous record; it is
* used to set the xl_prev of this record.
* 在WAL(buffer)中为给定大小的记录预留合适的空间。
* *StartPos设置为预留部分的开头，*EndPos设置为其结尾+1。
* *PrePtr设置为前一记录的开头;它用于设置该记录的xl_prev。
*
* This is the performance critical part of XLogInsert that must be serialized
* across backends. The rest can happen mostly in parallel. Try to keep this
* section as short as possible, insertpos_lck can be heavily contended on a
* busy system.
* 这是XLogInsert中与性能密切相关的部分，必须在后台进程之间序列执行。
* 其余的大部分可以同时发生。
* 尽量精简这部分的逻辑，insertpos_lck可以在繁忙的系统上存在激烈的竞争。
*
* NB: The space calculation here must match the code in CopyXLogRecordToWAL,
* where we actually copy the record to the reserved space.
* 注意:这里计算的空间必须与CopyXLogRecordToWAL()函数一致,
* 在CopyXLogRecordToWAL中会实际拷贝数据到预留空间中.
*/
static void
ReserveXLogInsertLocation(int size, XLogRecPtr *StartPos, XLogRecPtr *EndPos,
XLogRecPtr *PrevPtr)
{
XLogCtlInsert *Insert = &XLogCtl->Insert;//插入控制器
uint64 startbytepos;//开始位置
uint64 endbytepos;//结束位置
uint64 prevbytepos;//上一位置
size = MAXALIGN(size);//大小对齐
/* All (non xlog-switch) records should contain data. */
//除了xlog-switch外,所有的记录都应该包含数据.
Assert(size > SizeOfXLogRecord);
/*
* The duration the spinlock needs to be held is minimized by minimizing
* the calculations that have to be done while holding the lock. The
* current tip of reserved WAL is kept in CurrBytePos, as a byte position
* that only counts "usable" bytes in WAL, that is, it excludes all WAL
* page headers. The mapping between "usable" byte positions and physical
* positions (XLogRecPtrs) can be done outside the locked region, and
* because the usable byte position doesn't include any headers, reserving
* X bytes from WAL is almost as simple as "CurrBytePos += X".
* spinlock需要持有的时间通过最小化必须持有锁的计算逻辑达到最小化。
* 预留的WAL空间通过CurrBytePos变量(大小一个字节)保存，
* 它只计算WAL中的“可用”字节，也就是说，它排除了所有的WAL page header。
* “可用”字节位置和物理位置(XLogRecPtrs)之间的映射可以在锁定区域之外完成，
* 而且由于可用字节位置不包含任何header，从WAL预留X字节的大小几乎和“CurrBytePos += X”一样简单。
*/
SpinLockAcquire(&Insert->insertpos_lck);//申请锁
//开始位置
startbytepos = Insert->CurrBytePos;
//结束位置
endbytepos = startbytepos + size;
//上一位置
prevbytepos = Insert->PrevBytePos;
//调整控制器的相关变量
Insert->CurrBytePos = endbytepos;
Insert->PrevBytePos = startbytepos;
//释放锁
SpinLockRelease(&Insert->insertpos_lck);
//返回值
//计算开始/结束/上一位置偏移
*StartPos = XLogBytePosToRecPtr(startbytepos);
*EndPos = XLogBytePosToEndRecPtr(endbytepos);
*PrevPtr = XLogBytePosToRecPtr(prevbytepos);
/*
* Check that the conversions between "usable byte positions" and
* XLogRecPtrs work consistently in both directions.
* 检查双向转换之后的值是一致的.
*/
Assert(XLogRecPtrToBytePos(*StartPos) == startbytepos);
Assert(XLogRecPtrToBytePos(*EndPos) == endbytepos);
Assert(XLogRecPtrToBytePos(*PrevPtr) == prevbytepos);
}
/*
* Converts a "usable byte position" to XLogRecPtr. A usable byte position
* is the position starting from the beginning of WAL, excluding all WAL
* page headers.
* 将“可用字节位置”转换为XLogRecPtr。
* 可用字节位置是从WAL开始的位置，不包括所有WAL page header。
*/
static XLogRecPtr
XLogBytePosToRecPtr(uint64 bytepos)
{
uint64 fullsegs;
uint64 fullpages;
uint64 bytesleft;
uint32 seg_offset;
XLogRecPtr result;
fullsegs = bytepos / UsableBytesInSegment;
bytesleft = bytepos % UsableBytesInSegment;
if (bytesleft < XLOG_BLCKSZ - SizeOfXLogLongPHD)
{
//剩余的字节数 < XLOG_BLCKSZ - SizeOfXLogLongPHD
/* fits on first page of segment */
//填充在segment的第一个page中
seg_offset = bytesleft + SizeOfXLogLongPHD;
}
else
{
//剩余的字节数 >= XLOG_BLCKSZ - SizeOfXLogLongPHD
/* account for the first page on segment with long header */
//在segment中说明long header
seg_offset = XLOG_BLCKSZ;
bytesleft -= XLOG_BLCKSZ - SizeOfXLogLongPHD;
fullpages = bytesleft / UsableBytesInPage;
bytesleft = bytesleft % UsableBytesInPage;
seg_offset += fullpages * XLOG_BLCKSZ + bytesleft + SizeOfXLogShortPHD;
}
XLogSegNoOffsetToRecPtr(fullsegs, seg_offset, wal_segment_size, result);
return result;
}
/* The number of bytes in a WAL segment usable for WAL data. */
//WAL segment file中可用于WAL data的字节数(不包括page header)
static int UsableBytesInSegment;
CopyXLogRecordToWAL
CopyXLogRecordToWAL是XLogInsertRecord中的子过程,用于拷贝XLOG Record到WAL中的保留区域.
/*
* Subroutine of XLogInsertRecord. Copies a WAL record to an already-reserved
* area in the WAL.
* XLogInsertRecord中的子过程.
* 拷贝XLOG Record到WAL中的保留区域.
*/
static void
CopyXLogRecordToWAL(int write_len, bool isLogSwitch, XLogRecData *rdata,
XLogRecPtr StartPos, XLogRecPtr EndPos)
{
char *currpos;//当前指针位置
int freespace;//空闲空间
int written;//已写入的大小
XLogRecPtr CurrPos;//事务日志位置
XLogPageHeader pagehdr;//Page Header
/*
* Get a pointer to the right place in the right WAL buffer to start
* inserting to.
* 在合适的WAL buffer中获取指针用于确定插入的位置
*/
CurrPos = StartPos;//赋值为开始位置
currpos = GetXLogBuffer(CurrPos);//获取buffer指针
freespace = INSERT_FREESPACE(CurrPos);//获取空闲空间大小
/*
* there should be enough space for at least the first field (xl_tot_len)
* on this page.
* 在该页上最起码有第一个字段(xl_tot_len)的存储空间
*/
Assert(freespace >= sizeof(uint32));
/* Copy record data */
//拷贝记录数据
written = 0;
while (rdata != NULL)//循环
{
char *rdata_data = rdata->data;//指针
int rdata_len = rdata->len;//大小
while (rdata_len > freespace)//循环
{
/*
* Write what fits on this page, and continue on the next page.
* 该页能写多少就写多少,写不完就继续下一页.
*/
//确保最起码剩余SizeOfXLogShortPHD的头部数据存储空间
Assert(CurrPos % XLOG_BLCKSZ >= SizeOfXLogShortPHD || freespace == 0);
//内存拷贝
memcpy(currpos, rdata_data, freespace);
//指针调整
rdata_data += freespace;
//大小调整
rdata_len -= freespace;
//写入大小调整
written += freespace;
//当前位置调整
CurrPos += freespace;
/*
* Get pointer to beginning of next page, and set the xlp_rem_len
* in the page header. Set XLP_FIRST_IS_CONTRECORD.
* 获取下一页的开始指针,并在下一页的header中设置xlp_rem_len.
* 同时设置XLP_FIRST_IS_CONTRECORD标记.
*
* It's safe to set the contrecord flag and xlp_rem_len without a
* lock on the page. All the other flags were already set when the
* page was initialized, in AdvanceXLInsertBuffer, and we're the
* only backend that needs to set the contrecord flag.
* 就算不持有页锁,设置contrecord标记和xlp_rem_len也是安全的.
* 在页面初始化的时候,所有其他标记已通过AdvanceXLInsertBuffer函数初始化,
* 我们是需要设置contrecord标记的唯一一个后台进程,不会有其他进程了.
*/
currpos = GetXLogBuffer(CurrPos);//获取buffer
pagehdr = (XLogPageHeader) currpos;//获取page header
pagehdr->xlp_rem_len = write_len - written;//设置xlp_rem_len
pagehdr->xlp_info |= XLP_FIRST_IS_CONTRECORD;//设置标记
/* skip over the page header */
//跳过page header
if (XLogSegmentOffset(CurrPos, wal_segment_size) == 0)//第一个page
{
CurrPos += SizeOfXLogLongPHD;//Long Header
currpos += SizeOfXLogLongPHD;
}
else
{
CurrPos += SizeOfXLogShortPHD;//不是第一个page,Short Header
currpos += SizeOfXLogShortPHD;
}
freespace = INSERT_FREESPACE(CurrPos);//获取空闲空间
}
//再次验证
Assert(CurrPos % XLOG_BLCKSZ >= SizeOfXLogShortPHD || rdata_len == 0);
//内存拷贝(这时候rdata_len <= freespace)
memcpy(currpos, rdata_data, rdata_len);
currpos += rdata_len;//调整指针
CurrPos += rdata_len;//调整指针
freespace -= rdata_len;//减少空闲空间
written += rdata_len;//调整已写入大小
rdata = rdata->next;//下一批数据
}
Assert(written == write_len);//确保已写入 == 需写入大小
/*
* If this was an xlog-switch, it's not enough to write the switch record,
* we also have to consume all the remaining space in the WAL segment. We
* have already reserved that space, but we need to actually fill it.
* 如果是xlog-switch并且没有足够的空间写切换的记录,
* 这时候不得不消费WAL segment剩余的空间.
* 我们已经预留了空间,但需要执行实际的填充.
*/
if (isLogSwitch && XLogSegmentOffset(CurrPos, wal_segment_size) != 0)
{
/* An xlog-switch record doesn't contain any data besides the header */
//在header后,xlog-switch没有包含任何数据.
Assert(write_len == SizeOfXLogRecord);
/* Assert that we did reserve the right amount of space */
//验证预留了合适的空间
Assert(XLogSegmentOffset(EndPos, wal_segment_size) == 0);
/* Use up all the remaining space on the current page */
//在当前页面使用所有的剩余空间
CurrPos += freespace;
/*
* Cause all remaining pages in the segment to be flushed, leaving the
* XLog position where it should be, at the start of the next segment.
* We do this one page at a time, to make sure we don't deadlock
* against ourselves if wal_buffers < wal_segment_size.
* 由于该segment中所有剩余pages将被刷出,把XLog位置指向下一个segment的开始.
* 一个page我们只做一次,在wal_buffers < wal_segment_size的情况下,
* 确保我们自己不会出现死锁.
*/
while (CurrPos < EndPos)//循环
{
/*
* The minimal action to flush the page would be to call
* WALInsertLockUpdateInsertingAt(CurrPos) followed by
* AdvanceXLInsertBuffer(...). The page would be left initialized
* mostly to zeros, except for the page header (always the short
* variant, as this is never a segment's first page).
* 刷出page的最小化动作是:调用WALInsertLockUpdateInsertingAt(CurrPos)
* 然后接着调用AdvanceXLInsertBuffer(...).
* 除了page header(通常为short格式,除了segment的第一个page)外,其余部分均初始化为ascii 0.
*
* The large vistas of zeros are good for compressibility, but the
* headers interrupting them every XLOG_BLCKSZ (with values that
* differ from page to page) are not. The effect varies with
* compression tool, but bzip2 for instance compresses about an
* order of magnitude worse if those headers are left in place.
* 连续的ascii 0非常适合压缩,但每个page的头部数据(用于分隔page&page)把这些0隔开了.
* 这种效果随压缩工具的不同而不同，但是如果保留这些头文件，则bzip2的压缩效果会差一个数量级。
*
* Rather than complicating AdvanceXLInsertBuffer itself (which is
* called in heavily-loaded circumstances as well as this lightly-
* loaded one) with variant behavior, we just use GetXLogBuffer
* (which itself calls the two methods we need) to get the pointer
* and zero most of the page. Then we just zero the page header.
* 与其让AdvanceXLInsertBuffer本身(在重载环境和这个负载较轻的环境中调用)变得复杂，
* 不如使用GetXLogBuffer(调用了我们需要的两个方法)来初始化page(初始化为ascii 0)/
* 然后把page header设置为ascii 0.
*/
currpos = GetXLogBuffer(CurrPos);//获取buffer
MemSet(currpos, 0, SizeOfXLogShortPHD);//设置头部为ascii 0
CurrPos += XLOG_BLCKSZ;//修改指针
}
}
else
{
/* Align the end position, so that the next record starts aligned */
//对齐末尾位置,以便下一个记录可以从对齐的位置开始
CurrPos = MAXALIGN64(CurrPos);
}
if (CurrPos != EndPos)//验证
elog(PANIC, "space reserved for WAL record does not match what was written");
}
三、跟踪分析
测试脚本如下:
drop table t_wal_longtext;
create table t_wal_longtext(c1 int not null,c2 varchar(3000),c3 varchar(3000),c4 varchar(3000));
insert into t_wal_longtext(c1,c2,c3,c4)
select i,rpad('C2-'||i,3000,'2'),rpad('C3-'||i,3000,'3'),rpad('C4-'||i,3000,'4')
from generate_series(1,7) as i;
ReserveXLogInsertLocation
插入数据:
insert into t_wal_longtext(c1,c2,c3,c4) VALUES(8,'C2-8','C3-8','C4-8');
设置断点,进入ReserveXLogInsertLocation
(gdb) b ReserveXLogInsertLocation
Breakpoint 1 at 0x54d574: file xlog.c, line 1244.
(gdb) c
Continuing.
Breakpoint 1, ReserveXLogInsertLocation (size=74, StartPos=0x7ffebea9d768, EndPos=0x7ffebea9d760, PrevPtr=0x244f4c8)
at xlog.c:1244
1244 XLogCtlInsert *Insert = &XLogCtl->Insert;
(gdb)
输入参数:
size=74, 这是待插入XLOG Record的大小,其他三个为待设置的值.
继续执行.
对齐,74->80(要求为8的N倍,unit64占用8bytes,因此要求8的倍数)
(gdb) n
1249 size = MAXALIGN(size);
(gdb)
1252 Assert(size > SizeOfXLogRecord);
(gdb) p size
$1 = 80
(gdb)
查看插入控制器的信息,其中:
CurrBytePos = 5498377520,十六进制为0x147BA9530
PrevBytePos = 5498377464,十六进制为0x147BA94F8
RedoRecPtr = 5514382312,十六进制为0x148AECBE8 --> 对应pg_control中的Latest checkpoint's REDO location
(gdb) n
1264 SpinLockAcquire(&Insert->insertpos_lck);
(gdb)
1266 startbytepos = Insert->CurrBytePos;
(gdb) p *Insert
$2 = {insertpos_lck = 1 '\001', CurrBytePos = 5498377520, PrevBytePos = 5498377464, pad = '\000' <repeats 127 times>,
RedoRecPtr = 5514382312, forcePageWrites = false, fullPageWrites = true, exclusiveBackupState = EXCLUSIVE_BACKUP_NONE,
nonExclusiveBackups = 0, lastBackupStart = 0, WALInsertLocks = 0x7f97d1eeb100}
(gdb)
设置相应的值.
值得注意的是插入控制器Insert中的位置信息是不包括page header等信息,是纯粹可用的日志数据,因此数值要比WAL segment file的数值小.
(gdb) n
1267 endbytepos = startbytepos + size;
(gdb)
1268 prevbytepos = Insert->PrevBytePos;
(gdb)
1269 Insert->CurrBytePos = endbytepos;
(gdb)
1270 Insert->PrevBytePos = startbytepos;
(gdb)
1272 SpinLockRelease(&Insert->insertpos_lck);
(gdb)
如前所述,需要将“可用字节位置”转换为XLogRecPtr。
计算实际的开始/结束/上一位置.
StartPos = 5514538672,0x148B12EB0
EndPos = 5514538752,0x148B12F00
PrevPtr = 5514538616,0x148B12E78
(gdb) n
1274 *StartPos = XLogBytePosToRecPtr(startbytepos);
(gdb)
1275 *EndPos = XLogBytePosToEndRecPtr(endbytepos);
(gdb)
1276 *PrevPtr = XLogBytePosToRecPtr(prevbytepos);
(gdb)
1282 Assert(XLogRecPtrToBytePos(*StartPos) == startbytepos);
(gdb) p *StartPos
$4 = 5514538672
(gdb) p *EndPos
$5 = 5514538752
(gdb) p *PrevPtr
$6 = 5514538616
(gdb)
验证相互转换是没有问题的.
(gdb) n
1283 Assert(XLogRecPtrToBytePos(*EndPos) == endbytepos);
(gdb)
1284 Assert(XLogRecPtrToBytePos(*PrevPtr) == prevbytepos);
(gdb)
1285 }
(gdb)
XLogInsertRecord (rdata=0xf9cc70 <hdr_rdt>, fpw_lsn=5514538520, flags=1 '\001') at xlog.c:1072
1072 inserted = true;
(gdb)
DONE!
CopyXLogRecordToWAL-场景1:不跨WAL page
测试脚本如下:
insert into t_wal_longtext(c1,c2,c3,c4) VALUES(8,'C2-8','C3-8','C4-8');
继续上一条SQL的跟踪.
设置断点,进入CopyXLogRecordToWAL
(gdb) b CopyXLogRecordToWAL
Breakpoint 3 at 0x54dcdf: file xlog.c, line 1479.
(gdb) c
Continuing.
Breakpoint 3, CopyXLogRecordToWAL (write_len=74, isLogSwitch=false, rdata=0xf9cc70 <hdr_rdt>, StartPos=5514538672,
EndPos=5514538752) at xlog.c:1479
1479 CurrPos = StartPos;
(gdb)
输入参数:
write_len=74, --> 待写入大小
isLogSwitch=false, --> 是否日志切换(不需要)
rdata=0xf9cc70 <\hdr_rdt>, --> 需写入的数据地址
StartPos=5514538672, --> 开始位置
EndPos=5514538752 --> 结束位置
(gdb) n
1480 currpos = GetXLogBuffer(CurrPos);
(gdb)
在合适的WAL buffer中获取指针用于确定插入的位置.
进入函数GetXLogBuffer,输入参数ptr为5514538672,即开始位置.
(gdb) step
GetXLogBuffer (ptr=5514538672) at xlog.c:1854
1854 if (ptr / XLOG_BLCKSZ == cachedPage)
(gdb) p ptr / 8192 --> 取模
$7 = 673161
(gdb)
(gdb) p cachedPage
$8 = 673161
(gdb)
GetXLogBuffer->ptr / XLOG_BLCKSZ == cachedPage,进入相应的处理逻辑
注意:cachedPage是静态变量,具体在哪个地方赋值,后续需再行分析
(gdb) n
1856 Assert(((XLogPageHeader) cachedPos)->xlp_magic == XLOG_PAGE_MAGIC);
(gdb)
1857 Assert(((XLogPageHeader) cachedPos)->xlp_pageaddr == ptr - (ptr % XLOG_BLCKSZ));
(gdb)
1858 return cachedPos + ptr % XLOG_BLCKSZ;
GetXLogBuffer->cachedPos开头是XLogPageHeader结构体
(gdb) p *((XLogPageHeader) cachedPos)
$14 = {xlp_magic = 53400, xlp_info = 5, xlp_tli = 1, xlp_pageaddr = 5514534912, xlp_rem_len = 71}
(gdb)
(gdb) x/24bx (0x7f97d29fe000)
0x7f97d29fe000: 0x98 0xd0 0x05 0x00 0x01 0x00 0x00 0x00
0x7f97d29fe008: 0x00 0x20 0xb1 0x48 0x01 0x00 0x00 0x00
0x7f97d29fe010: 0x47 0x00 0x00 0x00 0x00 0x00 0x00 0x00
回到CopyXLogRecordToWAL,buffer的地址为0x7f97d29feeb0
(gdb) n
1945 }
(gdb)
CopyXLogRecordToWAL (write_len=74, isLogSwitch=false, rdata=0xf9cc70 <hdr_rdt>, StartPos=5514538672, EndPos=5514538752)
at xlog.c:1481
1481 freespace = INSERT_FREESPACE(CurrPos);
(gdb)
(gdb) p currpos
$16 = 0x7f97d29feeb0 ""
(gdb)
计算空闲空间,确保在该页上最起码有第一个字段(xl_tot_len)的存储空间(4字节).
(gdb) n
1487 Assert(freespace >= sizeof(uint32));
(gdb) p freespace
$21 = 4432
(gdb)
开始拷贝记录数据.
(gdb) n
1490 written = 0; --> 记录已写入的大小
(gdb)
1491 while (rdata != NULL)
rdata的分析详见第四部分,继续执行
(gdb) n
1493 char *rdata_data = rdata->data;
(gdb)
1494 int rdata_len = rdata->len;
(gdb)
1496 while (rdata_len > freespace)
(gdb) p rdata_len
$34 = 46
(gdb) p freespace
$35 = 4432
(gdb)
rdata_len < freespace,无需进入子循环.
再次进行验证没有问题,执行内存拷贝.
(gdb) n
1536 Assert(CurrPos % XLOG_BLCKSZ >= SizeOfXLogShortPHD || rdata_len == 0);
(gdb)
1537 memcpy(currpos, rdata_data, rdata_len);
(gdb)
1538 currpos += rdata_len;
(gdb)
1539 CurrPos += rdata_len;
(gdb)
1540 freespace -= rdata_len;
(gdb)
1541 written += rdata_len;
(gdb)
1543 rdata = rdata->next;
(gdb)
1491 while (rdata != NULL)
(gdb) p currpos
$36 = 0x7f97d29feede ""
(gdb) p CurrPos
$37 = 5514538718
(gdb) p freespace
$38 = 4386
(gdb) p written
$39 = 46
(gdb)
rdata共有四部分,继续写入第二/三/四部分.
...
1491 while (rdata != NULL)
(gdb)
1493 char *rdata_data = rdata->data;
(gdb)
1494 int rdata_len = rdata->len;
(gdb)
1496 while (rdata_len > freespace)
(gdb)
1536 Assert(CurrPos % XLOG_BLCKSZ >= SizeOfXLogShortPHD || rdata_len == 0);
(gdb)
1537 memcpy(currpos, rdata_data, rdata_len);
(gdb)
1538 currpos += rdata_len;
(gdb)
1539 CurrPos += rdata_len;
(gdb)
1540 freespace -= rdata_len;
(gdb)
1541 written += rdata_len;
(gdb)
1543 rdata = rdata->next;
(gdb)
1491 while (rdata != NULL)
(gdb)
完成写入74bytes
(gdb)
1545 Assert(written == write_len);
(gdb) p written
$40 = 74
(gdb)
无需执行日志切换的相关操作.
对齐CurrPos
(gdb) n
1552 if (isLogSwitch && XLogSegmentOffset(CurrPos, wal_segment_size) != 0)
(gdb)
1599 CurrPos = MAXALIGN64(CurrPos);
(gdb) p CurrPos
$41 = 5514538746
(gdb) n
1602 if (CurrPos != EndPos)
(gdb) p CurrPos
$42 = 5514538752
(gdb)
(gdb) p 5514538746 % 8
$44 = 2 --> 需补6个字节,5514538746 --> 5514538752
对齐后,CurrPos == EndPos,否则报错!
(gdb) p EndPos
$45 = 5514538752
结束调用
(gdb) n
1604 }
(gdb)
XLogInsertRecord (rdata=0xf9cc70 <hdr_rdt>, fpw_lsn=5514538520, flags=1 '\001') at xlog.c:1098
1098 if ((flags & XLOG_MARK_UNIMPORTANT) == 0)
(gdb)
DONE!
CopyXLogRecordToWAL-场景2:跨WAL page 后续再行分析
四、再论WAL Record
在内存中,WAL Record通过rdata存储,该变量其实是全局静态变量hdr_rdt,类型为XLogRecData,XLOG Record通过XLogRecData链表组织起来(这个设计很赞,写入无需理会结构,按链表逐个写数据即可).
rdata由4部分组成:
第一部分是XLogRecord + XLogRecordBlockHeader + XLogRecordDataHeaderShort,共46字节
第二部分是xl_heap_header,5个字节
第三部分是tuple data,20个字节
第四部分是xl_heap_insert,3个字节
------------------------------------------------------------------- 1
(gdb) p *rdata
$22 = {next = 0x244f2c0, data = 0x244f4c0 "J", len = 46}
(gdb) p *(XLogRecord *)rdata->data --> XLogRecord
$27 = {xl_tot_len = 74, xl_xid = 2268, xl_prev = 5514538616, xl_info = 0 '\000', xl_rmid = 10 '\n', xl_crc = 1158677949}
(gdb) p *(XLogRecordBlockHeader *)(0x244f4c0+24) --> XLogRecordBlockHeader
$29 = {id = 0 '\000', fork_flags = 32 ' ', data_length = 25}
(gdb) x/2bx (0x244f4c0+44) --> XLogRecordDataHeaderShort
0x244f4ec: 0xff 0x03
------------------------------------------------------------------- 2
(gdb) p *rdata->next
$23 = {next = 0x244f2d8, data = 0x7ffebea9d830 "\004", len = 5}
(gdb) p *(xl_heap_header *)rdata->next->data
$32 = {t_infomask2 = 4, t_infomask = 2050, t_hoff = 24 '\030'}
------------------------------------------------------------------- 3
(gdb) p *rdata->next->next
$24 = {next = 0x244f2a8, data = 0x24e6a2f "", len = 20}
(gdb) x/20bc 0x24e6a2f
0x24e6a2f: 0 '\000' 8 '\b' 0 '\000' 0 '\000' 0 '\000' 11 '\v' 67 'C' 50 '2'
0x24e6a37: 45 '-' 56 '8' 11 '\v' 67 'C' 51 '3' 45 '-' 56 '8' 11 '\v'
0x24e6a3f: 67 'C' 52 '4' 45 '-' 56 '8'
(gdb)
------------------------------------------------------------------- 4
(gdb) p *rdata->next->next->next
$25 = {next = 0x0, data = 0x7ffebea9d840 "\b", len = 3}
(gdb)
(gdb) p *(xl_heap_insert *)rdata->next->next->next->data
$33 = {offnum = 8, flags = 0 '\000'}
五、参考资料
PostgreSQL 源码解读（4）- 插入数据#3（heap_insert）
PostgreSQL 事务日志WAL结构浅析
PostgreSQL 源码解读（110）- WAL#6（Insert&WAL - XLogRecordAssemble记录组装函数）
PostgreSQL 源码解读（111）- WAL#7（Insert&WAL - XLogRecordAssemble-FPW）
PostgreSQL 源码解读（112）- WAL#8（XLogCtrl数据结构）
PG Source Code
