Support unlogged GiST index.

The reason this wasn't supported before was that GiST indexes need an
increasing sequence to detect concurrent page-splits. In a regular WAL-
logged GiST index, the LSN of the page-split record is used for that
purpose, and in a temporary index, we can get away with a backend-local
counter. Neither of those methods works for an unlogged relation.

To provide such an increasing sequence of numbers, create a "fake LSN"
counter that is saved and restored across shutdowns. On recovery, unlogged
relations are blown away, so the counter doesn't need to survive that
either.

Jeevan Chalke, based on discussions with Robert Haas, Tom Lane and me.

src/backend/access/gist/gist.c

@@ -16,6 +16,7 @@
 
 #include "access/genam.h"
 #include "access/gist_private.h"
+#include "access/heapam_xlog.h"
 #include "catalog/index.h"
 #include "catalog/pg_collation.h"
 #include "miscadmin.h"
@@ -71,9 +72,22 @@ createTempGistContext(void)
 Datum
 gistbuildempty(PG_FUNCTION_ARGS)
 {
-	ereport(ERROR,
-			(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-			 errmsg("unlogged GiST indexes are not supported")));
+	Relation	index = (Relation) PG_GETARG_POINTER(0);
+	Buffer		buffer;
+
+	/* Initialize the root page */
+	buffer = ReadBufferExtended(index, INIT_FORKNUM, P_NEW, RBM_NORMAL, NULL);
+	LockBuffer(buffer, BUFFER_LOCK_EXCLUSIVE);
+
+	/* Initialize and xlog buffer */
+	START_CRIT_SECTION();
+	GISTInitBuffer(buffer, F_LEAF);
+	MarkBufferDirty(buffer);
+	log_newpage_buffer(buffer);
+	END_CRIT_SECTION();
+
+	/* Unlock and release the buffer */
+	UnlockReleaseBuffer(buffer);
 
 	PG_RETURN_VOID();
 }
@@ -391,7 +405,7 @@ gistplacetopage(Relation rel, Size freespace, GISTSTATE *giststate,
 								   dist, oldrlink, oldnsn, leftchildbuf,
 								   markfollowright);
 		else
-			recptr = GetXLogRecPtrForTemp();
+			recptr = gistGetFakeLSN(rel);
 
 		for (ptr = dist; ptr; ptr = ptr->next)
 		{
@@ -448,7 +462,7 @@ gistplacetopage(Relation rel, Size freespace, GISTSTATE *giststate,
 		}
 		else
 		{
-			recptr = GetXLogRecPtrForTemp();
+			recptr = gistGetFakeLSN(rel);
 			PageSetLSN(page, recptr);
 		}
 

src/backend/access/gist/gistbuild.c

@@ -158,16 +158,6 @@ gistbuild(PG_FUNCTION_ARGS)
 		elog(ERROR, "index \"%s\" already contains data",
 			 RelationGetRelationName(index));
 
-	/*
-	 * We can't yet handle unlogged GiST indexes, because we depend on LSNs.
-	 * This is duplicative of an error in gistbuildempty, but we want to check
-	 * here so as to throw error before doing all the index-build work.
-	 */
-	if (heap->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED)
-		ereport(ERROR,
-				(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-				 errmsg("unlogged GiST indexes are not supported")));
-
 	/* no locking is needed */
 	buildstate.giststate = initGISTstate(index);
 
@@ -204,7 +194,7 @@ gistbuild(PG_FUNCTION_ARGS)
 		PageSetTLI(page, ThisTimeLineID);
 	}
 	else
-		PageSetLSN(page, GetXLogRecPtrForTemp());
+		PageSetLSN(page, gistGetFakeLSN(heap));
 
 	UnlockReleaseBuffer(buffer);
 

src/backend/access/gist/gistutil.c

@@ -798,16 +798,30 @@ gistoptions(PG_FUNCTION_ARGS)
 }
 
 /*
- * Temporary GiST indexes are not WAL-logged, but we need LSNs to detect
- * concurrent page splits anyway. GetXLogRecPtrForTemp() provides a fake
- * sequence of LSNs for that purpose. Each call generates an LSN that is
- * greater than any previous value returned by this function in the same
- * session.
+ * Temporary and unlogged GiST indexes are not WAL-logged, but we need LSNs
+ * to detect concurrent page splits anyway. This function provides a fake
+ * sequence of LSNs for that purpose.
  */
 XLogRecPtr
-GetXLogRecPtrForTemp(void)
+gistGetFakeLSN(Relation rel)
 {
 	static XLogRecPtr counter = 1;
-	counter++;
-	return counter;
+
+	if (rel->rd_rel->relpersistence == RELPERSISTENCE_TEMP)
+	{
+		/*
+		 * Temporary relations are only accessible in our session, so a
+		 * simple backend-local counter will do.
+		 */
+		return counter++;
+	}
+	else
+	{
+		/*
+		 * Unlogged relations are accessible from other backends, and survive
+		 * (clean) restarts. GetFakeLSNForUnloggedRel() handles that for us.
+		 */
+		Assert(rel->rd_rel->relpersistence == RELPERSISTENCE_UNLOGGED);
+		return GetFakeLSNForUnloggedRel();
+	}
 }

src/backend/access/gist/gistvacuum.c

@@ -238,7 +238,7 @@ gistbulkdelete(PG_FUNCTION_ARGS)
 					PageSetTLI(page, ThisTimeLineID);
 				}
 				else
-					PageSetLSN(page, GetXLogRecPtrForTemp());
+					PageSetLSN(page, gistGetFakeLSN(rel));
 
 				END_CRIT_SECTION();
 			}

src/backend/access/transam/xlog.c

@@ -391,6 +391,10 @@ typedef struct XLogCtlData
 	XLogRecPtr	asyncXactLSN;	/* LSN of newest async commit/abort */
 	XLogSegNo	lastRemovedSegNo; /* latest removed/recycled XLOG segment */
 
+	/* Fake LSN counter, for unlogged relations. Protected by ulsn_lck */
+	XLogRecPtr  unloggedLSN;
+	slock_t		ulsn_lck;
+
 	/* Protected by WALWriteLock: */
 	XLogCtlWrite Write;
 
@@ -3696,6 +3700,31 @@ GetSystemIdentifier(void)
 	return ControlFile->system_identifier;
 }
 
+/*
+ * Returns a fake LSN for unlogged relations.
+ *
+ * Each call generates an LSN that is greater than any previous value
+ * returned. The current counter value is saved and restored across clean
+ * shutdowns, but like unlogged relations, does not survive a crash. This can
+ * be used in lieu of real LSN values returned by XLogInsert, if you need an
+ * LSN-like increasing sequence of numbers without writing any WAL.
+ */
+XLogRecPtr
+GetFakeLSNForUnloggedRel(void)
+{
+	XLogRecPtr nextUnloggedLSN;
+
+	/* use volatile pointer to prevent code rearrangement */
+	volatile XLogCtlData *xlogctl = XLogCtl;
+
+	/* increment the unloggedLSN counter, need SpinLock */
+	SpinLockAcquire(&xlogctl->ulsn_lck);
+	nextUnloggedLSN = xlogctl->unloggedLSN++;
+	SpinLockRelease(&xlogctl->ulsn_lck);
+
+	return nextUnloggedLSN;
+}
+
 /*
  * Auto-tune the number of XLOG buffers.
  *
@@ -3844,6 +3873,7 @@ XLOGShmemInit(void)
 	XLogCtl->WalWriterSleeping = false;
 	XLogCtl->Insert.currpage = (XLogPageHeader) (XLogCtl->pages);
 	SpinLockInit(&XLogCtl->info_lck);
+	SpinLockInit(&XLogCtl->ulsn_lck);
 	InitSharedLatch(&XLogCtl->recoveryWakeupLatch);
 
 	/*
@@ -3989,6 +4019,7 @@ BootStrapXLOG(void)
 	ControlFile->time = checkPoint.time;
 	ControlFile->checkPoint = checkPoint.redo;
 	ControlFile->checkPointCopy = checkPoint;
+	ControlFile->unloggedLSN = 1;
 
 	/* Set important parameter values for use when replaying WAL */
 	ControlFile->MaxConnections = MaxConnections;
@@ -5032,6 +5063,16 @@ StartupXLOG(void)
 	XLogCtl->ckptXidEpoch = checkPoint.nextXidEpoch;
 	XLogCtl->ckptXid = checkPoint.nextXid;
 
+	/*
+	 * Initialize unlogged LSN. On a clean shutdown, it's restored from the
+	 * control file. On recovery, all unlogged relations are blown away, so
+	 * the unlogged LSN counter can be reset too.
+	 */
+	if (ControlFile->state == DB_SHUTDOWNED)
+		XLogCtl->unloggedLSN = ControlFile->unloggedLSN;
+	else
+		XLogCtl->unloggedLSN = 1;
+
 	/*
 	 * We must replay WAL entries using the same TimeLineID they were created
 	 * under, so temporarily adopt the TLI indicated by the checkpoint (see
@@ -6916,6 +6957,16 @@ CreateCheckPoint(int flags)
 	/* crash recovery should always recover to the end of WAL */
 	ControlFile->minRecoveryPoint = InvalidXLogRecPtr;
 	ControlFile->minRecoveryPointTLI = 0;
+
+	/*
+	 * Persist unloggedLSN value. It's reset on crash recovery, so this goes
+	 * unused on non-shutdown checkpoints, but seems useful to store it always
+	 * for debugging purposes.
+	 */
+	SpinLockAcquire(&XLogCtl->ulsn_lck);
+	ControlFile->unloggedLSN = XLogCtl->unloggedLSN;
+	SpinLockRelease(&XLogCtl->ulsn_lck);
+
 	UpdateControlFile();
 	LWLockRelease(ControlFileLock);