Change WAL-logging scheme for multixacts to be more like regular

transaction IDs, rather than like subtrans; in particular, the information
now survives a database restart.  Per previous discussion, this is
essential for PITR log shipping and for 2PC.

Author: tglsfdc

doc/src/sgml/ref/pg_resetxlog.sgml

@@ -1,5 +1,5 @@
 <!--
-$PostgreSQL: pgsql/doc/src/sgml/ref/pg_resetxlog.sgml,v 1.10 2005/04/28 21:47:10 tgl Exp $
+$PostgreSQL: pgsql/doc/src/sgml/ref/pg_resetxlog.sgml,v 1.11 2005/06/08 15:50:21 tgl Exp $
 PostgreSQL documentation
 -->
 
@@ -23,6 +23,7 @@ PostgreSQL documentation
    <arg> -o <replaceable class="parameter">oid</replaceable> </arg>
    <arg> -x <replaceable class="parameter">xid</replaceable> </arg>
    <arg> -m <replaceable class="parameter">mxid</replaceable> </arg>
+   <arg> -O <replaceable class="parameter">mxoff</replaceable> </arg>
    <arg> -l <replaceable class="parameter">timelineid</replaceable>,<replaceable class="parameter">fileid</replaceable>,<replaceable class="parameter">seg</replaceable> </arg>
    <arg choice="plain"><replaceable>datadir</replaceable></arg>
   </cmdsynopsis>
@@ -32,8 +33,8 @@ PostgreSQL documentation
   <title>Description</title>
   <para>
    <command>pg_resetxlog</command> clears the write-ahead log (WAL) and
-   optionally resets some other control information (stored in the
-   <filename>pg_control</> file).  This function is sometimes needed
+   optionally resets some other control information stored in the
+   <filename>pg_control</> file.  This function is sometimes needed
    if these files have become corrupted.  It should be used only as a
    last resort, when the server will not start due to such corruption.
   </para>
@@ -60,8 +61,9 @@ PostgreSQL documentation
    by specifying the <literal>-f</> (force) switch.  In this case plausible
    values will be substituted for the missing data.  Most of the fields can be
    expected to match, but manual assistance may be needed for the next OID,
-   next transaction ID, WAL starting address, and database locale fields.
-   The first three of these can be set using the switches discussed below.
+   next transaction ID, next multi-transaction ID and offset,
+   WAL starting address, and database locale fields.
+   The first five of these can be set using the switches discussed below.
    <command>pg_resetxlog</command>'s own environment is the source for its
    guess at the locale fields; take care that <envar>LANG</> and so forth
    match the environment that <command>initdb</> was run in.
@@ -74,9 +76,10 @@ PostgreSQL documentation
   </para>
 
   <para>
-   The <literal>-o</>, <literal>-x</>, <literal>-m</>, and <literal>-l</>
+   The <literal>-o</>, <literal>-x</>, <literal>-m</>, <literal>-O</>,
+   and <literal>-l</>
    switches allow the next OID, next transaction ID, next multi-transaction
-   ID, and WAL starting address values to
+   ID, next multi-transaction offset, and WAL starting address values to
    be set manually.  These are only needed when
    <command>pg_resetxlog</command> is unable to determine appropriate values
    by reading <filename>pg_control</>.  Safe values may be determined as
@@ -108,6 +111,17 @@ PostgreSQL documentation
      </para>
     </listitem>
 
+    <listitem>
+     <para>
+      A safe value for the next multi-transaction offset (<literal>-O</>)
+      may be determined by looking for the numerically largest
+      file name in the directory <filename>pg_multixact/members</> under the
+      data directory, adding one, and then multiplying by 65536.  As above,
+      the file names are in hexadecimal, so the easiest way to do this is to
+      specify the switch value in hexadecimal and add four zeroes.
+     </para>
+    </listitem>
+
     <listitem>
      <para>
       The WAL starting address (<literal>-l</>) should be

src/backend/access/heap/heapam.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.193 2005/06/06 20:22:56 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.194 2005/06/08 15:50:21 tgl Exp $
  *
  *
  * INTERFACE ROUTINES
@@ -2219,6 +2219,8 @@ heap_lock_tuple(Relation relation, HeapTuple tuple, Buffer *buffer,
 	 * Else the same IDs might be re-used after a crash, which would be
 	 * disastrous if this page made it to disk before the crash.  Essentially
 	 * we have to enforce the WAL log-before-data rule even in this case.
+	 * (Also, in a PITR log-shipping or 2PC environment, we have to have XLOG
+	 * entries for everything anyway.)
 	 */
 	if (!relation->rd_istemp)
 	{
@@ -2228,6 +2230,8 @@ heap_lock_tuple(Relation relation, HeapTuple tuple, Buffer *buffer,
 
 		xlrec.target.node = relation->rd_node;
 		xlrec.target.tid = tuple->t_self;
+		xlrec.locking_xid = xid;
+		xlrec.xid_is_mxact = ((new_infomask & HEAP_XMAX_IS_MULTI) != 0);
 		xlrec.shared_lock = (mode == LockTupleShared);
 		rdata[0].data = (char *) &xlrec;
 		rdata[0].len = SizeOfHeapLock;
@@ -2900,17 +2904,18 @@ heap_xlog_lock(XLogRecPtr lsn, XLogRecord *record)
 
 	htup = (HeapTupleHeader) PageGetItem(page, lp);
 
-	/*
-	 * Presently, we don't bother to restore the locked state, but
-	 * just set the XMAX_INVALID bit.
-	 */
 	htup->t_infomask &= ~(HEAP_XMAX_COMMITTED |
 						  HEAP_XMAX_INVALID |
 						  HEAP_XMAX_IS_MULTI |
 						  HEAP_IS_LOCKED |
 						  HEAP_MOVED);
-	htup->t_infomask |= HEAP_XMAX_INVALID;
-	HeapTupleHeaderSetXmax(htup, record->xl_xid);
+	if (xlrec->xid_is_mxact)
+		htup->t_infomask |= HEAP_XMAX_IS_MULTI;
+	if (xlrec->shared_lock)
+		htup->t_infomask |= HEAP_XMAX_SHARED_LOCK;
+	else
+		htup->t_infomask |= HEAP_XMAX_EXCL_LOCK;
+	HeapTupleHeaderSetXmax(htup, xlrec->locking_xid);
 	HeapTupleHeaderSetCmax(htup, FirstCommandId);
 	/* Make sure there is no forward chain link in t_ctid */
 	htup->t_ctid = xlrec->target.tid;
@@ -3010,6 +3015,11 @@ heap_desc(char *buf, uint8 xl_info, char *rec)
 			strcat(buf, "shared_lock: ");
 		else
 			strcat(buf, "exclusive_lock: ");
+		if (xlrec->xid_is_mxact)
+			strcat(buf, "mxid ");
+		else
+			strcat(buf, "xid ");
+		sprintf(buf + strlen(buf), "%u ", xlrec->locking_xid);
 		out_target(buf, &(xlrec->target));
 	}
 	else