PostgreSQL 源码解读（118）- MVCC#3(Tuple可见性判断)

本节介绍了PostgreSQL判断元组可见性的实现逻辑,重点解析了HeapTupleSatisfiesMVCC函数的处理过程。

一、数据结构

宏定义

//事务ID,无符号32位整型typedef uint32 TransactionId;/* ---------------- *      Special transaction ID values *      系统保留的事务ID * * BootstrapTransactionId is the XID for "bootstrap" operations, and * FrozenTransactionId is used for very old tuples.  Both should * always be considered valid. * * FirstNormalTransactionId is the first "normal" transaction id. * Note: if you need to change it, you must change pg_class.h as well. * ---------------- */#define InvalidTransactionId        ((TransactionId) 0)#define BootstrapTransactionId      ((TransactionId) 1)#define FrozenTransactionId         ((TransactionId) 2)#define FirstNormalTransactionId    ((TransactionId) 3)#define MaxTransactionId            ((TransactionId) 0xFFFFFFFF)/* ---------------- *      transaction ID manipulation macros * ---------------- */#define TransactionIdIsValid(xid)       ((xid) != InvalidTransactionId)#define TransactionIdIsNormal(xid)      ((xid) >= FirstNormalTransactionId)#define TransactionIdEquals(id1, id2)   ((id1) == (id2))#define TransactionIdStore(xid, dest)   (*(dest) = (xid))#define StoreInvalidTransactionId(dest) (*(dest) = InvalidTransactionId)/* advance a transaction ID variable, handling wraparound correctly */#define TransactionIdAdvance(dest)  \    do { \        (dest)++; \        if ((dest) < FirstNormalTransactionId) \            (dest) = FirstNormalTransactionId; \    } while(0)/* back up a transaction ID variable, handling wraparound correctly */#define TransactionIdRetreat(dest)  \    do { \        (dest)--; \    } while ((dest) < FirstNormalTransactionId)/* compare two XIDs already known to be normal; this is a macro for speed *///比较两个已知是常规事务的XIDs;宏定义是为了性能考虑.#define NormalTransactionIdPrecedes(id1, id2) \    (AssertMacro(TransactionIdIsNormal(id1) && TransactionIdIsNormal(id2)), \    (int32) ((id1) - (id2)) < 0)/* compare two XIDs already known to be normal; this is a macro for speed */#define NormalTransactionIdFollows(id1, id2) \    (AssertMacro(TransactionIdIsNormal(id1) && TransactionIdIsNormal(id2)), \    (int32) ((id1) - (id2)) > 0)/* * HeapTupleHeaderGetRawXmin returns the "raw" xmin field, which is the xid * originally used to insert the tuple.  However, the tuple might actually * be frozen (via HeapTupleHeaderSetXminFrozen) in which case the tuple's xmin * is visible to every snapshot.  Prior to PostgreSQL 9.4, we actually changed * the xmin to FrozenTransactionId, and that value may still be encountered * on disk. * HeapTupleHeaderGetRawXmin返回"原始"xmin字段，这是最初用于插入元组的xid。 * 但是，实际上元组实可能被冻结(通过HeapTupleHeaderSetXminFrozen)， *   在这种情况下，元组可对每个快照都可见。 * 在PostgreSQL 9.4之前，我们实际上将xmin更改为FrozenTransactionId，该值仍然可能在磁盘上遇到。 */#define HeapTupleHeaderGetRawXmin(tup) \( \    (tup)->t_choice.t_heap.t_xmin \)#define HeapTupleHeaderXminCommitted(tup) \( \    ((tup)->t_infomask & HEAP_XMIN_COMMITTED) != 0 \)/* * HeapTupleHeaderGetRawXmax gets you the raw Xmax field.  To find out the Xid * that updated a tuple, you might need to resolve the MultiXactId if certain * bits are set.  HeapTupleHeaderGetUpdateXid checks those bits and takes care * to resolve the MultiXactId if necessary.  This might involve multixact I/O, * so it should only be used if absolutely necessary. * HeapTupleHeaderGetRawXmax返回原始的Xmax字段. * 搜索更新元组的事务Xid,如设定了特定的标记位则可能需要解决MultiXactId. * HeapTupleHeaderGetUpdateXid检查这些标记位并在需要时解决MultiXactId问题. * 这可能会引起multixact I/O,如非必要不要使用. */#define HeapTupleHeaderGetUpdateXid(tup) \( \    (!((tup)->t_infomask & HEAP_XMAX_INVALID) && \     ((tup)->t_infomask & HEAP_XMAX_IS_MULTI) && \     !((tup)->t_infomask & HEAP_XMAX_LOCK_ONLY)) ? \        HeapTupleGetUpdateXid(tup) \    : \        HeapTupleHeaderGetRawXmax(tup) \)#define HeapTupleHeaderGetRawXmax(tup) \( \    (tup)->t_choice.t_heap.t_xmax \)/* * A tuple is only locked (i.e. not updated by its Xmax) if the * HEAP_XMAX_LOCK_ONLY bit is set; or, for pg_upgrade's sake, if the Xmax is * not a multi and the EXCL_LOCK bit is set. * 如HEAP_XMAX_LOCK_ONLY标记位设置,则表示仅仅只是锁定了tuple(比如并非更新了Xmax字段). * 或者,对于pg_upgrade升级,如Xmax并非HEAP_XMAX_IS_MULTI而且EXCL_LOCK标记位设置,也是表达这个意思. * * See also HeapTupleHeaderIsOnlyLocked, which also checks for a possible * aborted updater transaction. * 可同时参考HeapTupleHeaderIsOnlyLocked,该函数也检查了可能已回滚的更新事务. * * Beware of multiple evaluations of the argument. * 注意对参数的多种分析. */#define HEAP_XMAX_IS_LOCKED_ONLY(infomask) \    (((infomask) & HEAP_XMAX_LOCK_ONLY) || \     (((infomask) & (HEAP_XMAX_IS_MULTI | HEAP_LOCK_MASK)) == HEAP_XMAX_EXCL_LOCK))

二、源码解读

全局变量snapshot->satisfies函数在初始化该全局变量已设置为HeapTupleSatisfiesMVCC.

static SnapshotData CurrentSnapshotData = {HeapTupleSatisfiesMVCC};

HeapTupleSatisfiesMVCC - 如Tuple对于给定的MVCC快照可见,则返回T

/* * HeapTupleSatisfiesMVCC *      True iff heap tuple is valid for the given MVCC snapshot. * HeapTupleSatisfiesMVCC -- 如Tuple对于给定的MVCC快照可见,则返回T * *  Here, we consider the effects of: *      all transactions committed as of the time of the given snapshot *      previous commands of this transaction * 在这里，我们考虑的是: *     在给定快照之前的事务命令执行时提交的所有事务 * *  Does _not_ include: *      transactions shown as in-progress by the snapshot *      transactions started after the snapshot was taken *      changes made by the current command * 但不包括: *      快照显示仍在进行中的事务. *      快照之后才启动的事务. *      * * Notice that here, we will not update the tuple status hint bits if the * inserting/deleting transaction is still running according to our snapshot, * even if in reality it's committed or aborted by now.  This is intentional. * Checking the true transaction state would require access to high-traffic * shared data structures, creating contention we'd rather do without, and it * would not change the result of our visibility check anyway.  The hint bits * will be updated by the first visitor that has a snapshot new enough to see * the inserting/deleting transaction as done.  In the meantime, the cost of * leaving the hint bits unset is basically that each HeapTupleSatisfiesMVCC * call will need to run TransactionIdIsCurrentTransactionId in addition to * XidInMVCCSnapshot (but it would have to do the latter anyway).  In the old * coding where we tried to set the hint bits as soon as possible, we instead * did TransactionIdIsInProgress in each call --- to no avail, as long as the * inserting/deleting transaction was still running --- which was more cycles * and more contention on the PGXACT array. * 在这里需要注意的是,如果插入/删除事务仍然在快照范围内运行， *   那么我们将不会更新元组状态提示位，即使它现在已经提交或中止。这是做法内部约定的. * 检查真正的事务状态需要访问高频访问的共享数据结构， *   这就产生了我们不希望看到的争用，而且它也不会改变可见性检查的结果。 * 第一个访问者将更新提示位，该访问者的快照足够新，可以看到插入/删除事务已经完成。 * 与此同时，不设置提示位的代价是，除了XidInMVCCSnapshot之外， *   每个HeapTupleSatisfiesMVCC调用都需要运行TransactionIdIsCurrentTransactionId *   (无论如何都必须执行后者)。 * 在旧的编码中，我们试图尽快地设置提示位， *   而不是在每次调用中执行TransactionIdIsInProgress-- *      只要插入/删除事务仍在运行,这些工作看起来都是徒劳的 *   --导致PGXACT数组上更多的循环和争用。 */boolHeapTupleSatisfiesMVCC(HeapTuple htup, Snapshot snapshot,                       Buffer buffer){    HeapTupleHeader tuple = htup->t_data;//Tuple    Assert(ItemPointerIsValid(&htup->t_self));    Assert(htup->t_tableOid != InvalidOid);    if (!HeapTupleHeaderXminCommitted(tuple))    {        //A.xmin事务未提交(HEAP_XMIN_COMMITTED标记未设置)        if (HeapTupleHeaderXminInvalid(tuple))            //xmin = 0,此tuple不可见            return false;        /* Used by pre-9.0 binary upgrades */        //用于9.0前二进制可执行文件的升级,HEAP_MOVED_OFF & HEAP_MOVED_IN已不再使用        if (tuple->t_infomask & HEAP_MOVED_OFF)        {            TransactionId xvac = HeapTupleHeaderGetXvac(tuple);            if (TransactionIdIsCurrentTransactionId(xvac))                return false;            if (!XidInMVCCSnapshot(xvac, snapshot))            {                if (TransactionIdDidCommit(xvac))                {                    SetHintBits(tuple, buffer, HEAP_XMIN_INVALID,                                InvalidTransactionId);                    return false;                }                SetHintBits(tuple, buffer, HEAP_XMIN_COMMITTED,                            InvalidTransactionId);            }        }        /* Used by pre-9.0 binary upgrades */        //同上        else if (tuple->t_infomask & HEAP_MOVED_IN)        {            TransactionId xvac = HeapTupleHeaderGetXvac(tuple);            if (!TransactionIdIsCurrentTransactionId(xvac))            {                if (XidInMVCCSnapshot(xvac, snapshot))                    return false;                if (TransactionIdDidCommit(xvac))                    SetHintBits(tuple, buffer, HEAP_XMIN_COMMITTED,                                InvalidTransactionId);                else                {                    SetHintBits(tuple, buffer, HEAP_XMIN_INVALID,                                InvalidTransactionId);                    return false;                }            }        }        else if (TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmin(tuple)))        {            //A.xmin事务未提交            //A1.xmin对应的事务是当前事务            if (HeapTupleHeaderGetCmin(tuple) >= snapshot->curcid)                //A1-1.tuple.cmin >= 快照的cid,说明插入在扫描开始前,返回F                return false;   /* inserted after scan started */            if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid */                //A1-2.xmax无效,返回T                return true;            if (HEAP_XMAX_IS_LOCKED_ONLY(tuple->t_infomask))    /* not deleter */                //A1-3.XMAX仅仅只是锁定而已,返回T                return true;            if (tuple->t_infomask & HEAP_XMAX_IS_MULTI)            {                //xmax is a MultiXactId                 TransactionId xmax;                //获取更新的Xid                xmax = HeapTupleGetUpdateXid(tuple);                /* not LOCKED_ONLY, so it has to have an xmax */                //验证xmax是有效的                Assert(TransactionIdIsValid(xmax));                /* updating subtransaction must have aborted */                //执行更新的子事务必须已回滚(why?因为前提是xmin事务是未提交的,因此xmax必须已回滚)                if (!TransactionIdIsCurrentTransactionId(xmax))                    //正在执行删除的事务不是本事务,返回T                    return true;                else if (HeapTupleHeaderGetCmax(tuple) >= snapshot->curcid)                    //本事务执行删除:删除命令在快照之后,返回T                    return true;    /* updated after scan started */                else                    //本事务执行删除:删除命令在快照之前,返回F                    return false;   /* updated before scan started */            }            if (!TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmax(tuple)))            {                //A1-4.tuple.xmax对应的事务不是本事务                /* deleting subtransaction must have aborted */                //删除子事务已终止(why?因为前提是xmin事务是未提交的,因此xmax必须已回滚)                SetHintBits(tuple, buffer, HEAP_XMAX_INVALID,                            InvalidTransactionId);                return true;            }            if (HeapTupleHeaderGetCmax(tuple) >= snapshot->curcid)                //A1-5.删除在快照之后,返回T                return true;    /* deleted after scan started */            else                //A1-6.删在快照之前,返回F                return false;   /* deleted before scan started */        }        else if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmin(tuple), snapshot))            //A.xmin事务未提交;            //A2.xmin在当前快照中(非本事务) --> 不可见            return false;        else if (TransactionIdDidCommit(HeapTupleHeaderGetRawXmin(tuple)))            //A.xmin事务未提交;            //A3.查询事务日志clog,Xmin事务已提交,则设置标志位,可见判断逻辑在后续执行            SetHintBits(tuple, buffer, HEAP_XMIN_COMMITTED,                        HeapTupleHeaderGetRawXmin(tuple));        else        {            /* it must have aborted or crashed */            //A4.以上各种情况都不符合,事务回滚或者在执行事务过程中crash了            //设置标志位,返回F            SetHintBits(tuple, buffer, HEAP_XMIN_INVALID,                        InvalidTransactionId);            return false;        }    }    else    {        /* xmin is committed, but maybe not according to our snapshot */        //B.xmin事务已提交,但可能不属于该快照        if (!HeapTupleHeaderXminFrozen(tuple) &&            XidInMVCCSnapshot(HeapTupleHeaderGetRawXmin(tuple), snapshot))            return false;       /* treat as still in progress */    }    /* by here, the inserting transaction has committed */    //C.Xmin事务已提交    if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid or aborted */        //C1.xmax无效(HEAP_XMAX_INVALID),则返回T        return true;    if (HEAP_XMAX_IS_LOCKED_ONLY(tuple->t_infomask))        //C2.xmax仅仅只是lock而已,返回T        return true;    if (tuple->t_infomask & HEAP_XMAX_IS_MULTI)    {        //xmax is a MultiXactId         TransactionId xmax;        /* already checked above */        //再次检查        Assert(!HEAP_XMAX_IS_LOCKED_ONLY(tuple->t_infomask));        //获取xmax        xmax = HeapTupleGetUpdateXid(tuple);        /* not LOCKED_ONLY, so it has to have an xmax */        Assert(TransactionIdIsValid(xmax));        if (TransactionIdIsCurrentTransactionId(xmax))        {            //xmax为当前事务id            if (HeapTupleHeaderGetCmax(tuple) >= snapshot->curcid)                //快照之后执行删除,返回T                return true;    /* deleted after scan started */            else                //快照之前执行删除,返回F                return false;   /* deleted before scan started */        }        if (XidInMVCCSnapshot(xmax, snapshot))            //xmax仍在进行中,返回T            return true;        if (TransactionIdDidCommit(xmax))            //xmax事务通过查询clog日志,已确认提交,则返回F            return false;       /* updating transaction committed */        /* it must have aborted or crashed */        //xmax回滚或者执行期间crash,返回T        return true;    }    if (!(tuple->t_infomask & HEAP_XMAX_COMMITTED))    {        //C3.xmax未提交        if (TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmax(tuple)))        {            //C3-1:本事务            if (HeapTupleHeaderGetCmax(tuple) >= snapshot->curcid)                return true;    /* deleted after scan started */            else                return false;   /* deleted before scan started */        }        if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmax(tuple), snapshot))            //C3-2:非本事务            //xmax事务仍在进行中            return true;        if (!TransactionIdDidCommit(HeapTupleHeaderGetRawXmax(tuple)))        {            /* it must have aborted or crashed */            //C3-3查询clog日志,xmax事务未提交(回滚或crash),设置标记            SetHintBits(tuple, buffer, HEAP_XMAX_INVALID,                        InvalidTransactionId);            //可见            return true;        }        /* xmax transaction committed */        //以上判断均不成立,则可认为事务已提交,设置标记        SetHintBits(tuple, buffer, HEAP_XMAX_COMMITTED,                    HeapTupleHeaderGetRawXmax(tuple));    }    else    {        /* xmax is committed, but maybe not according to our snapshot */        //C4.xmax已提交,但快照指示该事务仍在进行中        if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmax(tuple), snapshot))            //仍在处理中,返回T            return true;        /* treat as still in progress */    }    /* xmax transaction committed */    //C5.xmax已提交,返回F    return false;}

三、跟踪分析

创建数据表,插入数据,删除其中一行数据,提交,更新其中一条数据,不提交,执行查询.
session 1

11:55:00 (xdb@[local]:5432)testdb=# drop table t_mvcc2;ERROR:  table "t_mvcc2" does not exist11:55:01 (xdb@[local]:5432)testdb=# create table t_mvcc2(c1 int not null,c2  varchar(40),c3 varchar(40));CREATE TABLE11:55:02 (xdb@[local]:5432)testdb=# 11:55:02 (xdb@[local]:5432)testdb=# insert into t_mvcc2(c1,c2,c3) values(1,'C2-1','C3-1');INSERT 0 111:55:02 (xdb@[local]:5432)testdb=# insert into t_mvcc2(c1,c2,c3) values(2,'C2-2','C3-2');INSERT 0 111:55:03 (xdb@[local]:5432)testdb=# 11:55:38 (xdb@[local]:5432)testdb=# begin;BEGIN11:55:38 (xdb@[local]:5432)testdb=#* 11:55:38 (xdb@[local]:5432)testdb=#* delete from t_mvcc2 where c1 = 1;DELETE 111:55:38 (xdb@[local]:5432)testdb=#* 11:55:38 (xdb@[local]:5432)testdb=#* commit;COMMIT11:55:38 (xdb@[local]:5432)testdb=# 11:55:38 (xdb@[local]:5432)testdb=# begin;BEGIN11:55:38 (xdb@[local]:5432)testdb=#* 11:55:38 (xdb@[local]:5432)testdb=#* update t_mvcc2 set c2 = 'C2#'||substr(c2,4,40) where c1 = 2;UPDATE 111:55:38 (xdb@[local]:5432)testdb=#* 12:19:17 (xdb@[local]:5432)testdb=#* select txid_current(); txid_current --------------         2363(1 row)

另外启动session 2,查询数据表t_mvcc2

11:58:45 (xdb@[local]:5432)testdb=# select lp,lp_off,lp_flags,t_xmin,t_xmax,t_field3 as t_cid,t_ctid,t_infomask2,t_infomask from heap_page_items(get_raw_page('t_mvcc2',0)); lp | lp_off | lp_flags | t_xmin | t_xmax | t_cid | t_ctid | t_infomask2 | t_infomask ----+--------+----------+--------+--------+-------+--------+-------------+------------  1 |   8152 |        1 |   2360 |   2362 |     0 | (0,1)  |        8195 |       1282  2 |   8112 |        1 |   2361 |   2363 |     0 | (0,3)  |       16387 |        258  3 |   8072 |        1 |   2363 |      0 |     0 | (0,3)  |       32771 |      10242(3 rows)11:59:38 (xdb@[local]:5432)testdb=# select * from t_mvcc2;

启动gdb,设置断点

(gdb) b HeapTupleSatisfiesMVCCBreakpoint 1 at 0xa93125: file tqual.c, line 966.(gdb) cContinuing.Breakpoint 1, HeapTupleSatisfiesMVCC (htup=0x1985d18, snapshot=0x1a06358, buffer=69) at tqual.c:966966     HeapTupleHeader tuple = htup->t_data;(gdb)

查看调用栈

(gdb) bt#0  HeapTupleSatisfiesMVCC (htup=0x1985d18, snapshot=0x1a06358, buffer=69) at tqual.c:966#1  0x00000000004de959 in heap_hot_search_buffer (tid=0x1985d1c, relation=0x7f002f73e2b0, buffer=69, snapshot=0x1a06358,     heapTuple=0x1985d18, all_dead=0x7ffc56ed8862, first_call=true) at heapam.c:2127#2  0x00000000004fd12e in index_fetch_heap (scan=0x1985cb8) at indexam.c:608#3  0x00000000004fd35d in index_getnext (scan=0x1985cb8, direction=ForwardScanDirection) at indexam.c:691#4  0x00000000004fb910 in systable_getnext (sysscan=0x19868f8) at genam.c:425#5  0x0000000000a1cb63 in SearchCatCacheMiss (cache=0x19de480, nkeys=1, hashValue=1281076841, hashIndex=105, v1=42610,     v2=0, v3=0, v4=0) at catcache.c:1386#6  0x0000000000a1ca10 in SearchCatCacheInternal (cache=0x19de480, nkeys=1, v1=42610, v2=0, v3=0, v4=0) at catcache.c:1317#7  0x0000000000a1c6fe in SearchCatCache1 (cache=0x19de480, v1=42610) at catcache.c:1185#8  0x0000000000a37543 in SearchSysCache1 (cacheId=50, key1=42610) at syscache.c:1119#9  0x0000000000a6ae60 in check_enable_rls (relid=42610, checkAsUser=0, noError=false) at rls.c:66#10 0x000000000087b286 in get_row_security_policies (root=0x1985dd0, rte=0x1985ee8, rt_index=1,     securityQuals=0x7ffc56ed8cc0, withCheckOptions=0x7ffc56ed8cb8, hasRowSecurity=0x7ffc56ed8cb7,     hasSubLinks=0x7ffc56ed8cb6) at rowsecurity.c:133#11 0x0000000000875e1c in fireRIRrules (parsetree=0x1985dd0, activeRIRs=0x0) at rewriteHandler.c:1904#12 0x0000000000878b23 in QueryRewrite (parsetree=0x1985dd0) at rewriteHandler.c:3712#13 0x00000000008c64d7 in pg_rewrite_query (query=0x1985dd0) at postgres.c:782#14 0x00000000008c634e in pg_analyze_and_rewrite (parsetree=0x1985c20, query_string=0x1984ec8 "select * from t_mvcc2;",     paramTypes=0x0, numParams=0, queryEnv=0x0) at postgres.c:698#15 0x00000000008c6993 in exec_simple_query (query_string=0x1984ec8 "select * from t_mvcc2;") at postgres.c:1070#16 0x00000000008cae70 in PostgresMain (argc=1, argv=0x19b0dc8, dbname=0x19b0c30 "testdb", username=0x1981ba8 "xdb")    at postgres.c:4182#17 0x000000000082642b in BackendRun (port=0x19a6c00) at postmaster.c:4361#18 0x0000000000825b8f in BackendStartup (port=0x19a6c00) at postmaster.c:4033#19 0x0000000000821f1c in ServerLoop () at postmaster.c:1706#20 0x00000000008217b4 in PostmasterMain (argc=1, argv=0x197fb60) at postmaster.c:1379---Type  to continue, or q  to quit---#21 0x00000000007488ef in main (argc=1, argv=0x197fb60) at main.c:228(gdb)

第一个Tuple
输入参数,主要是tuple/snapshot/buffer

(gdb) p *snapshot --> 快照信息$3 = {satisfies = 0xa9310d , xmin = 2363, xmax = 2363, xip = 0x0, xcnt = 0, subxip = 0x0,   subxcnt = 0, suboverflowed = false, takenDuringRecovery = false, copied = true, curcid = 0, speculativeToken = 0,   active_count = 0, regd_count = 1, ph_node = {first_child = 0x0, next_sibling = 0x0,     prev_or_parent = 0xf9bfa0 }, whenTaken = 0, lsn = 0}

获取tuple
t_infomask = 2313,十六进制为0x0909,即HEAP_XMAX_INVALID | HEAP_XMIN_COMMITTED | HEAP_HASOID | HEAP_HASNULL

(gdb) n968     Assert(ItemPointerIsValid(&htup->t_self));(gdb) p tuple$1 = (HeapTupleHeader) 0x7f0002da7a60(gdb) p *tuple$2 = {t_choice = {t_heap = {t_xmin = 2359, t_xmax = 0, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {datum_len_ = 2359,       datum_typmod = 0, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 12}, ip_posid = 6}, t_infomask2 = 33,   t_infomask = 2313, t_hoff = 32 ' ', t_bits = 0x7f0002da7a77 "\377\377\377?"}(gdb)

判断tuple.xmin是否已提交,按上一步的t_infomask标记,该事务已提交,进入相应逻辑

(gdb) n969     Assert(htup->t_tableOid != InvalidOid);(gdb) 971     if (!HeapTupleHeaderXminCommitted(tuple))(gdb)

判断:B.xmin事务已提交,但可能不属于该快照
不符合此条件,继续执行

(gdb) n1074                XidInMVCCSnapshot(HeapTupleHeaderGetRawXmin(tuple), snapshot))(gdb) 1073            if (!HeapTupleHeaderXminFrozen(tuple) &&(gdb) 1080        if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid or aborted */(gdb)

判断是否HEAP_XMAX_INVALID,按t_infomask标记,符合条件,返回T

(gdb) 1080        if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid or aborted */(gdb) n1081            return true;(gdb) 1148    }(gdb)

第二个Tuple
接下来是第二个Tuple

(gdb) cContinuing.Breakpoint 1, HeapTupleSatisfiesMVCC (htup=0x1985d18, snapshot=0x1a06358, buffer=155) at tqual.c:966966     HeapTupleHeader tuple = htup->t_data;(gdb)

获取tuple并查看
t_infomask = 2313,十六进制为0x0909,即HEAP_XMAX_INVALID | HEAP_XMIN_COMMITTED | HEAP_HASOID | HEAP_HASNULL

(gdb) p *htup$4 = {t_len = 172, t_self = {ip_blkid = {bi_hi = 0, bi_lo = 1}, ip_posid = 40}, t_tableOid = 1247, t_data = 0x7f0002e52800}(gdb) n968     Assert(ItemPointerIsValid(&htup->t_self));(gdb) p tuple$5 = (HeapTupleHeader) 0x7f0002e52800(gdb) p *tuple$6 = {t_choice = {t_heap = {t_xmin = 1, t_xmax = 0, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {datum_len_ = 1,       datum_typmod = 0, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 1}, ip_posid = 40}, t_infomask2 = 30,   t_infomask = 2313, t_hoff = 32 ' ', t_bits = 0x7f0002e52817 "\377\377\377\a"}(gdb)

与第一个Tuple类似,下面查看第三个Tupel
第三个Tuple
获取tuple并查看
t_infomask = 10505,十六进制值为0x2909,即HEAP_UPDATED | HEAP_XMAX_INVALID | HEAP_XMIN_COMMITTED | HEAP_HASOID | HEAP_HASNULL

(gdb) cContinuing.Breakpoint 1, HeapTupleSatisfiesMVCC (htup=0x1985d18, snapshot=0xf9bf60 , buffer=4) at tqual.c:966966     HeapTupleHeader tuple = htup->t_data;(gdb) n968     Assert(ItemPointerIsValid(&htup->t_self));(gdb) 969     Assert(htup->t_tableOid != InvalidOid);(gdb) p *tuple$7 = {t_choice = {t_heap = {t_xmin = 2117, t_xmax = 0, t_field3 = {t_cid = 7, t_xvac = 7}}, t_datum = {datum_len_ = 2117,       datum_typmod = 0, datum_typeid = 7}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 1}, ip_posid = 58}, t_infomask2 = 33,   t_infomask = 10505, t_hoff = 32 ' ', t_bits = 0x7f0002d25aa7 "\377\377\377?"}(gdb)

第三个Tuple,类似于第一/二个Tuple,仍返回T

(gdb) n971     if (!HeapTupleHeaderXminCommitted(tuple))(gdb) 1073            if (!HeapTupleHeaderXminFrozen(tuple) &&(gdb) 1074                XidInMVCCSnapshot(HeapTupleHeaderGetRawXmin(tuple), snapshot))(gdb) 1073            if (!HeapTupleHeaderXminFrozen(tuple) &&(gdb) 1080        if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid or aborted */(gdb) 1081            return true;(gdb) 1148    }

第四/五/六个Tuple
均有HEAP_XMAX_INVALID标志,不作展开

(gdb) cContinuing.Breakpoint 1, HeapTupleSatisfiesMVCC (htup=0x1985d18, snapshot=0x1a06358, buffer=44) at tqual.c:966966     HeapTupleHeader tuple = htup->t_data;(gdb) n968     Assert(ItemPointerIsValid(&htup->t_self));(gdb) 969     Assert(htup->t_tableOid != InvalidOid);(gdb) p *tuple$8 = {t_choice = {t_heap = {t_xmin = 1, t_xmax = 0, t_field3 = {t_cid = 75, t_xvac = 75}}, t_datum = {datum_len_ = 1,       datum_typmod = 0, datum_typeid = 75}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 21}, ip_posid = 1}, t_infomask2 = 24,   t_infomask = 2305, t_hoff = 32 ' ', t_bits = 0x7f0002d76307 "\377\377\017"}(gdb) ...(gdb) p *tuple$9 = {t_choice = {t_heap = {t_xmin = 1, t_xmax = 0, t_field3 = {t_cid = 75, t_xvac = 75}}, t_datum = {datum_len_ = 1,       datum_typmod = 0, datum_typeid = 75}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 1}, ip_posid = 19}, t_infomask2 = 20,   t_infomask = 2307, t_hoff = 32 ' ', t_bits = 0x7f0002d7368f "\377\377\003"}...(gdb) p *tuple$10 = {t_choice = {t_heap = {t_xmin = 1, t_xmax = 0, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {datum_len_ = 1,       datum_typmod = 0, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 0}, ip_posid = 1}, t_infomask2 = 4,   t_infomask = 2313, t_hoff = 32 ' ', t_bits = 0x7f0002ee632f "\003"}

第七个Tuple
t_infomask = 1282,0x0502,即HEAP_XMAX_COMMITTED | HEAP_XMIN_COMMITTED | HEAP_HASVARWIDTH

(gdb) cContinuing.Breakpoint 1, HeapTupleSatisfiesMVCC (htup=0x7ffc56ed8920, snapshot=0x1a062c0, buffer=206) at tqual.c:966966     HeapTupleHeader tuple = htup->t_data;(gdb) n968     Assert(ItemPointerIsValid(&htup->t_self));(gdb) 969     Assert(htup->t_tableOid != InvalidOid);(gdb) p *tuple$11 = {t_choice = {t_heap = {t_xmin = 2360, t_xmax = 2362, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {      datum_len_ = 2360, datum_typmod = 2362, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 0},     ip_posid = 1}, t_infomask2 = 8195, t_infomask = 1282, t_hoff = 24 '\030', t_bits = 0x7f0002eba36f ""}(gdb)

这是被"deleted"的tuple,2362事务已提交,不可见

1086        if (tuple->t_infomask & HEAP_XMAX_IS_MULTI)(gdb) 1113        if (!(tuple->t_infomask & HEAP_XMAX_COMMITTED))(gdb) 1141            if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmax(tuple), snapshot))(gdb) n1147        return false;(gdb)

第八个Tuple
t_infomask = 258,0x0102,即HEAP_XMIN_COMMITTED | HEAP_HASVARWIDTH

(gdb) p *tuple$12 = {t_choice = {t_heap = {t_xmin = 2361, t_xmax = 2363, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {      datum_len_ = 2361, datum_typmod = 2363, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 0},     ip_posid = 3}, t_infomask2 = 16387, t_infomask = 258, t_hoff = 24 '\030', t_bits = 0x7f0002eba347 ""}(gdb)

这是正在update的tuple,应可见

1080        if (tuple->t_infomask & HEAP_XMAX_INVALID)  /* xid invalid or aborted */(gdb) 1083        if (HEAP_XMAX_IS_LOCKED_ONLY(tuple->t_infomask))(gdb) 1086        if (tuple->t_infomask & HEAP_XMAX_IS_MULTI)(gdb) 1113        if (!(tuple->t_infomask & HEAP_XMAX_COMMITTED))(gdb) 1115            if (TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmax(tuple)))  --> 非当前事务(gdb) 1123            if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmax(tuple), snapshot)) --> 事务快照标志该事务进行中(gdb) 1124                return true; --> 可见(gdb)

第九个Tuple
t_infomask = 10242,0x2802,即HEAP_UPDATED | HEAP_XMAX_INVALID | HEAP_HASVARWIDTH

(gdb) p *tuple$13 = {t_choice = {t_heap = {t_xmin = 2363, t_xmax = 0, t_field3 = {t_cid = 0, t_xvac = 0}}, t_datum = {datum_len_ = 2363,       datum_typmod = 0, datum_typeid = 0}}, t_ctid = {ip_blkid = {bi_hi = 0, bi_lo = 0}, ip_posid = 3},   t_infomask2 = 32771, t_infomask = 10242, t_hoff = 24 '\030', t_bits = 0x7f0002eba31f ""}(gdb)

这是update操作新生成的tuple,不可见

(gdb) n971     if (!HeapTupleHeaderXminCommitted(tuple)) --> xmin未提交(gdb) 973         if (HeapTupleHeaderXminInvalid(tuple))(gdb) 977         if (tuple->t_infomask & HEAP_MOVED_OFF)(gdb) 996         else if (tuple->t_infomask & HEAP_MOVED_IN)(gdb) 1015            else if (TransactionIdIsCurrentTransactionId(HeapTupleHeaderGetRawXmin(tuple))) --> 非当前事务(gdb) 1057            else if (XidInMVCCSnapshot(HeapTupleHeaderGetRawXmin(tuple), snapshot)) --> xmin事务处于活动中(gdb) 1058                return false; --> 不可见(gdb)

查询结果

11:59:38 (xdb@[local]:5432)testdb=# select * from t_mvcc2; c1 |  c2  |  c3  ----+------+------  2 | C2-2 | C3-2(1 row)

DONE!

四、参考资料

PG Source Code