public inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH 2/2] make LogwrtResult atomic
26+ messages / 3 participants
[nested] [flat]

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH v5 1/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 192 +++++++++++++++---------------
 src/include/port/atomics.h        |  29 +++++
 src/tools/pgindent/typedefs.list  |   1 +
 3 files changed, 123 insertions(+), 99 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 221e4cb34f..225ac8bc2d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -382,16 +382,14 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *
  * LogwrtRqst indicates a byte position that we need to write and/or fsync
  * the log up to (all records before that point must be written or fsynced).
- * LogwrtResult indicates the byte positions we have already written/fsynced.
+ * LogWriteResult, LogFlushResult indicate the byte positions we have already
+ * written/fsynced.
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -414,18 +412,18 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;		/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;		/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
 	XLogRecPtr	Flush;			/* last byte + 1 to flush */
 } XLogwrtRqst;
 
-typedef struct XLogwrtResult
-{
-	XLogRecPtr	Write;			/* last byte + 1 written out */
-	XLogRecPtr	Flush;			/* last byte + 1 flushed */
-} XLogwrtResult;
-
 /*
  * Inserting to WAL is protected by a small fixed number of WAL insertion
  * locks. To insert to the WAL, you must hold one of the locks - it doesn't
@@ -577,6 +575,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult; /* uses atomics */
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -594,12 +593,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -770,10 +763,11 @@ static ControlFileData *ControlFile = NULL;
 static int	UsableBytesInSegment;
 
 /*
- * Private, possibly out-of-date copy of shared LogwrtResult.
+ * Private, possibly out-of-date copy of shared XLogCtl->LogwrtResult.
  * See discussion above.
  */
-static XLogwrtResult LogwrtResult = {0, 0};
+static XLogRecPtr LogWriteResult = 0;
+static XLogRecPtr LogFlushResult = 0;
 
 /*
  * Codes indicating where we got a WAL file from during recovery, or where
@@ -1201,8 +1195,6 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 	}
 
@@ -2174,6 +2166,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, TimeLineID tli, bool opportunistic)
 	 * Now that we have the lock, check if someone initialized the page
 	 * already.
 	 */
+	LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 	while (upto >= XLogCtl->InitializedUpTo || opportunistic)
 	{
 		nextidx = XLogRecPtrToBufIdx(XLogCtl->InitializedUpTo);
@@ -2184,7 +2177,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, TimeLineID tli, bool opportunistic)
 		 * already written out.
 		 */
 		OldPageRqstPtr = XLogCtl->xlblocks[nextidx];
-		if (LogwrtResult.Write < OldPageRqstPtr)
+		if (LogWriteResult < OldPageRqstPtr)
 		{
 			/*
 			 * Nope, got work to do. If we just want to pre-initialize as much
@@ -2193,18 +2186,18 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, TimeLineID tli, bool opportunistic)
 			if (opportunistic)
 				break;
 
-			/* Before waiting, get info_lck and update LogwrtResult */
+			/* Advance shared memory write request position */
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
-			 * Now that we have an up-to-date LogwrtResult value, see if we
-			 * still need to write it or if someone else already did.
+			 * Before waiting, update LogWriteResult and see if we still need
+			 * to write it or if someone else already did.
 			 */
-			if (LogwrtResult.Write < OldPageRqstPtr)
+			LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+			if (LogWriteResult < OldPageRqstPtr)
 			{
 				/*
 				 * Must acquire write lock. Release WALBufMappingLock first,
@@ -2218,8 +2211,8 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, TimeLineID tli, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
-				if (LogwrtResult.Write >= OldPageRqstPtr)
+				LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+				if (LogWriteResult >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
 					LWLockRelease(WALWriteLock);
@@ -2476,7 +2469,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2496,9 +2489,9 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 	 * consider writing.  Begin at the buffer containing the next unwritten
 	 * page, or last partially written page.
 	 */
-	curridx = XLogRecPtrToBufIdx(LogwrtResult.Write);
+	curridx = XLogRecPtrToBufIdx(LogWriteResult);
 
-	while (LogwrtResult.Write < WriteRqst.Write)
+	while (LogWriteResult < WriteRqst.Write)
 	{
 		/*
 		 * Make sure we're not ahead of the insert process.  This could happen
@@ -2507,16 +2500,16 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		 */
 		XLogRecPtr	EndPtr = XLogCtl->xlblocks[curridx];
 
-		if (LogwrtResult.Write >= EndPtr)
+		if (LogWriteResult >= EndPtr)
 			elog(PANIC, "xlog write request %X/%X is past end of log %X/%X",
-				 LSN_FORMAT_ARGS(LogwrtResult.Write),
+				 LSN_FORMAT_ARGS(LogWriteResult),
 				 LSN_FORMAT_ARGS(EndPtr));
 
-		/* Advance LogwrtResult.Write to end of current buffer page */
-		LogwrtResult.Write = EndPtr;
-		ispartialpage = WriteRqst.Write < LogwrtResult.Write;
+		/* Advance LogWriteResult to end of current buffer page */
+		LogWriteResult = EndPtr;
+		ispartialpage = WriteRqst.Write < LogWriteResult;
 
-		if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
+		if (!XLByteInPrevSeg(LogWriteResult, openLogSegNo,
 							 wal_segment_size))
 		{
 			/*
@@ -2526,7 +2519,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 			Assert(npages == 0);
 			if (openLogFile >= 0)
 				XLogFileClose();
-			XLByteToPrevSeg(LogwrtResult.Write, openLogSegNo,
+			XLByteToPrevSeg(LogWriteResult, openLogSegNo,
 							wal_segment_size);
 			openLogTLI = tli;
 
@@ -2538,7 +2531,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		/* Make sure we have the current logfile open */
 		if (openLogFile < 0)
 		{
-			XLByteToPrevSeg(LogwrtResult.Write, openLogSegNo,
+			XLByteToPrevSeg(LogWriteResult, openLogSegNo,
 							wal_segment_size);
 			openLogTLI = tli;
 			openLogFile = XLogFileOpen(openLogSegNo, tli);
@@ -2550,7 +2543,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		{
 			/* first of group */
 			startidx = curridx;
-			startoffset = XLogSegmentOffset(LogwrtResult.Write - XLOG_BLCKSZ,
+			startoffset = XLogSegmentOffset(LogWriteResult - XLOG_BLCKSZ,
 											wal_segment_size);
 		}
 		npages++;
@@ -2561,7 +2554,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		 * contiguous in memory), or if we are at the end of the logfile
 		 * segment.
 		 */
-		last_iteration = WriteRqst.Write <= LogwrtResult.Write;
+		last_iteration = WriteRqst.Write <= LogWriteResult;
 
 		finishing_seg = !ispartialpage &&
 			(startoffset + npages * XLOG_BLCKSZ) >= wal_segment_size;
@@ -2652,13 +2645,13 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 				/* signal that we need to wakeup walsenders later */
 				WalSndWakeupRequest();
 
-				LogwrtResult.Flush = LogwrtResult.Write;	/* end of page */
+				LogFlushResult = LogWriteResult;	/* end of page */
 
 				if (XLogArchivingActive())
 					XLogArchiveNotifySeg(openLogSegNo, tli);
 
 				XLogCtl->lastSegSwitchTime = (pg_time_t) time(NULL);
-				XLogCtl->lastSegSwitchLSN = LogwrtResult.Flush;
+				XLogCtl->lastSegSwitchLSN = LogFlushResult;
 
 				/*
 				 * Request a checkpoint if we've consumed too much xlog since
@@ -2679,7 +2672,7 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		if (ispartialpage)
 		{
 			/* Only asked to write a partial page */
-			LogwrtResult.Write = WriteRqst.Write;
+			LogWriteResult = WriteRqst.Write;
 			break;
 		}
 		curridx = NextBufIdx(curridx);
@@ -2691,12 +2684,14 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 
 	Assert(npages == 0);
 
+	/* Publish current write result position */
+	pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogWriteResult);
+
 	/*
 	 * If asked to flush, do so
 	 */
-	if (LogwrtResult.Flush < WriteRqst.Flush &&
-		LogwrtResult.Flush < LogwrtResult.Write)
-
+	if (LogFlushResult < WriteRqst.Flush &&
+		LogFlushResult < LogWriteResult)
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2707,12 +2702,12 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 			sync_method != SYNC_METHOD_OPEN_DSYNC)
 		{
 			if (openLogFile >= 0 &&
-				!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
+				!XLByteInPrevSeg(LogWriteResult, openLogSegNo,
 								 wal_segment_size))
 				XLogFileClose();
 			if (openLogFile < 0)
 			{
-				XLByteToPrevSeg(LogwrtResult.Write, openLogSegNo,
+				XLByteToPrevSeg(LogWriteResult, openLogSegNo,
 								wal_segment_size);
 				openLogTLI = tli;
 				openLogFile = XLogFileOpen(openLogSegNo, tli);
@@ -2725,23 +2720,23 @@ XLogWrite(XLogwrtRqst WriteRqst, TimeLineID tli, bool flexible)
 		/* signal that we need to wakeup walsenders later */
 		WalSndWakeupRequest();
 
-		LogwrtResult.Flush = LogwrtResult.Write;
+		LogFlushResult = LogWriteResult;
 	}
 
+	/* Publish current flush result position */
+	pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogFlushResult);
+
 	/*
-	 * Update shared-memory status
-	 *
-	 * We make sure that the shared 'request' values do not fall behind the
+	 * Make sure that the shared 'request' values do not fall behind the
 	 * 'result' values.  This is not absolutely essential, but it saves some
 	 * code in a couple of places.
 	 */
 	{
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
-		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
-			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
-		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
-			XLogCtl->LogwrtRqst.Flush = LogwrtResult.Flush;
+		if (XLogCtl->LogwrtRqst.Write < LogWriteResult)
+			XLogCtl->LogwrtRqst.Write = LogWriteResult;
+		if (XLogCtl->LogwrtRqst.Flush < LogFlushResult)
+			XLogCtl->LogwrtRqst.Flush = LogFlushResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 	}
 }
@@ -2757,8 +2752,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2775,7 +2770,7 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 		WriteRqstPtr -= WriteRqstPtr % XLOG_BLCKSZ;
 
 		/* if we have already flushed that far, we're done */
-		if (WriteRqstPtr <= LogwrtResult.Flush)
+		if (WriteRqstPtr <= LogFlushResult)
 			return;
 	}
 
@@ -2931,15 +2926,15 @@ XLogFlush(XLogRecPtr record)
 	}
 
 	/* Quick exit if already known flushed */
-	if (record <= LogwrtResult.Flush)
+	if (record <= LogFlushResult)
 		return;
 
 #ifdef WAL_DEBUG
 	if (XLOG_DEBUG)
 		elog(LOG, "xlog flush request %X/%X; write %X/%X; flush %X/%X",
 			 LSN_FORMAT_ARGS(record),
-			 LSN_FORMAT_ARGS(LogwrtResult.Write),
-			 LSN_FORMAT_ARGS(LogwrtResult.Flush));
+			 LSN_FORMAT_ARGS(LogWriteResult),
+			 LSN_FORMAT_ARGS(LogFlushResult));
 #endif
 
 	START_CRIT_SECTION();
@@ -2948,8 +2943,8 @@ XLogFlush(XLogRecPtr record)
 	 * Since fsync is usually a horribly expensive operation, we try to
 	 * piggyback as much data as we can on each fsync: if we see any more data
 	 * entered into the xlog buffer, we'll write and fsync that too, so that
-	 * the final value of LogwrtResult.Flush is as large as possible. This
-	 * gives us some chance of avoiding another fsync immediately after.
+	 * the final value of LogFlushResult is as large as possible. This gives
+	 * us some chance of avoiding another fsync immediately after.
 	 */
 
 	/* initialize to given target; may increase below */
@@ -2967,11 +2962,11 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
-		if (record <= LogwrtResult.Flush)
+		LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
+		if (record <= LogFlushResult)
 			break;
 
 		/*
@@ -2998,8 +2993,8 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
-		if (record <= LogwrtResult.Flush)
+		LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
+		if (record <= LogFlushResult)
 		{
 			LWLockRelease(WALWriteLock);
 			break;
@@ -3069,11 +3064,11 @@ XLogFlush(XLogRecPtr record)
 	 * calls from bufmgr.c are not within critical sections and so we will not
 	 * force a restart for a bad LSN on a data page.
 	 */
-	if (LogwrtResult.Flush < record)
+	if (LogFlushResult < record)
 		elog(ERROR,
 			 "xlog flush request %X/%X is not satisfied --- flushed only to %X/%X",
 			 LSN_FORMAT_ARGS(record),
-			 LSN_FORMAT_ARGS(LogwrtResult.Flush));
+			 LSN_FORMAT_ARGS(LogFlushResult));
 }
 
 /*
@@ -3122,7 +3117,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3130,8 +3124,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
-	if (WriteRqst.Write <= LogwrtResult.Flush)
+	LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
+	if (WriteRqst.Write <= LogFlushResult)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3143,11 +3139,12 @@ XLogBackgroundFlush(void)
 	 * holding an open file handle to a logfile that's no longer in use,
 	 * preventing the file from being deleted.
 	 */
-	if (WriteRqst.Write <= LogwrtResult.Flush)
+	if (WriteRqst.Write <= LogFlushResult)
 	{
 		if (openLogFile >= 0)
 		{
-			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
+			LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+			if (!XLByteInPrevSeg(LogWriteResult, openLogSegNo,
 								 wal_segment_size))
 			{
 				XLogFileClose();
@@ -3162,7 +3159,7 @@ XLogBackgroundFlush(void)
 	 */
 	now = GetCurrentTimestamp();
 	flushbytes =
-		WriteRqst.Write / XLOG_BLCKSZ - LogwrtResult.Flush / XLOG_BLCKSZ;
+		WriteRqst.Write / XLOG_BLCKSZ - LogFlushResult / XLOG_BLCKSZ;
 
 	if (WalWriterFlushAfter == 0 || lastflush == 0)
 	{
@@ -3197,8 +3194,8 @@ XLogBackgroundFlush(void)
 		elog(LOG, "xlog bg flush request write %X/%X; flush: %X/%X, current is write %X/%X; flush %X/%X",
 			 LSN_FORMAT_ARGS(WriteRqst.Write),
 			 LSN_FORMAT_ARGS(WriteRqst.Flush),
-			 LSN_FORMAT_ARGS(LogwrtResult.Write),
-			 LSN_FORMAT_ARGS(LogwrtResult.Flush));
+			 LSN_FORMAT_ARGS(LogWriteResult),
+			 LSN_FORMAT_ARGS(LogFlushResult));
 #endif
 
 	START_CRIT_SECTION();
@@ -3206,9 +3203,10 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	if (WriteRqst.Write > LogwrtResult.Write ||
-		WriteRqst.Flush > LogwrtResult.Flush)
+	LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
+	if (WriteRqst.Write > LogWriteResult ||
+		WriteRqst.Flush > LogFlushResult)
 	{
 		XLogWrite(WriteRqst, insertTLI, flexible);
 	}
@@ -3290,16 +3288,14 @@ XLogNeedsFlush(XLogRecPtr record)
 	}
 
 	/* Quick exit if already known flushed */
-	if (record <= LogwrtResult.Flush)
+	if (record <= LogFlushResult)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
-	if (record <= LogwrtResult.Flush)
+	if (record <= LogFlushResult)
 		return false;
 
 	return true;
@@ -8086,9 +8082,11 @@ StartupXLOG(void)
 		XLogCtl->InitializedUpTo = EndOfLog;
 	}
 
-	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
+	LogWriteResult = LogFlushResult = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	/* XXX OK to write without WALWriteLock? */
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8750,9 +8748,7 @@ GetFlushRecPtr(TimeLineID *insertTLI)
 {
 	Assert(XLogCtl->SharedRecoveryState == RECOVERY_STATE_DONE);
 
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogFlushResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/*
 	 * If we're writing and flushing WAL, the time line can't be changing,
@@ -8761,7 +8757,7 @@ GetFlushRecPtr(TimeLineID *insertTLI)
 	if (insertTLI)
 		*insertTLI = XLogCtl->InsertTimeLineID;
 
-	return LogwrtResult.Flush;
+	return LogFlushResult;
 }
 
 /*
@@ -12041,11 +12037,9 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogWriteResult = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
-	return LogwrtResult.Write;
+	return LogWriteResult;
 }
 
 /*
diff --git a/src/include/port/atomics.h b/src/include/port/atomics.h
index 856338f161..4ff9ce9864 100644
--- a/src/include/port/atomics.h
+++ b/src/include/port/atomics.h
@@ -519,6 +519,35 @@ pg_atomic_sub_fetch_u64(volatile pg_atomic_uint64 *ptr, int64 sub_)
 	return pg_atomic_sub_fetch_u64_impl(ptr, sub_);
 }
 
+/*
+ * Monotonically advance the given variable using only atomic operations until
+ * it's at least the target value.
+ *
+ * Full barrier semantics (even when value is unchanged).
+ */
+static inline void
+pg_atomic_monotonic_advance_u64(volatile pg_atomic_uint64 *ptr, uint64 target_)
+{
+	uint64		currval;
+
+#ifndef PG_HAVE_ATOMIC_U64_SIMULATION
+	AssertPointerAlignment(ptr, 8);
+#endif
+
+	currval = pg_atomic_read_u64(ptr);
+	if (currval >= target_)
+	{
+		pg_memory_barrier();
+		return;
+	}
+
+	while (currval < target_)
+	{
+		if (pg_atomic_compare_exchange_u64(ptr, &currval, target_))
+			break;
+	}
+}
+
 #undef INSIDE_ATOMICS_H
 
 #endif							/* ATOMICS_H */
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index da6ac8ed83..edc9bfa5ce 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -2943,6 +2943,7 @@ XLogRecordBuffer
 XLogRedoAction
 XLogSegNo
 XLogSource
+XLogwrtAtomic
 XLogwrtResult
 XLogwrtRqst
 XPVIV
-- 
2.30.2


--fl2voahem2nkfart
Content-Type: text/x-diff; charset=utf-8
Content-Disposition: attachment; filename="v5-0002-more-atomics.patch"



^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* [PATCH 2/2] make LogwrtResult atomic
@ 2021-02-02 17:03  Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 26+ messages in thread

From: Alvaro Herrera @ 2021-02-02 17:03 UTC (permalink / raw)

---
 src/backend/access/transam/xlog.c | 66 +++++++++++++++----------------
 1 file changed, 32 insertions(+), 34 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..10802bd56f 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -405,12 +405,9 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  * These structs are identical but are declared separately to indicate their
  * slightly different functions.
  *
- * To read XLogCtl->LogwrtResult, you must hold either info_lck or
- * WALWriteLock.  To update it, you need to hold both locks.  The point of
- * this arrangement is that the value can be examined by code that already
- * holds WALWriteLock without needing to grab info_lck as well.  In addition
- * to the shared variable, each backend has a private copy of LogwrtResult,
- * which is updated when convenient.
+ * XLogCtl->LogwrtResult is read and written using atomic operations.
+ * In addition to the shared variable, each backend has a private copy of
+ * LogwrtResult, each member of which is separately updated when convenient.
  *
  * The request bookkeeping is simpler: there is a shared XLogCtl->LogwrtRqst
  * (protected by info_lck), but we don't need to cache any copies of it.
@@ -433,6 +430,12 @@ static XLogRecPtr RedoStartLSN = InvalidXLogRecPtr;
  *----------
  */
 
+typedef struct XLogwrtAtomic
+{
+	pg_atomic_uint64 Write;			/* last byte + 1 of write position */
+	pg_atomic_uint64 Flush;			/* last byte + 1 of flush position */
+} XLogwrtAtomic;
+
 typedef struct XLogwrtRqst
 {
 	XLogRecPtr	Write;			/* last byte + 1 to write out */
@@ -596,6 +599,7 @@ typedef struct XLogCtlData
 {
 	XLogCtlInsert Insert;
 
+	XLogwrtAtomic LogwrtResult;
 	/* Protected by info_lck: */
 	XLogwrtRqst LogwrtRqst;
 	XLogRecPtr	RedoRecPtr;		/* a recent copy of Insert->RedoRecPtr */
@@ -613,12 +617,6 @@ typedef struct XLogCtlData
 	pg_time_t	lastSegSwitchTime;
 	XLogRecPtr	lastSegSwitchLSN;
 
-	/*
-	 * Protected by info_lck and WALWriteLock (you must hold either lock to
-	 * read it, but both to update)
-	 */
-	XLogwrtResult LogwrtResult;
-
 	/*
 	 * Latest initialized page in the cache (last byte position + 1).
 	 *
@@ -1172,9 +1170,10 @@ XLogInsertRecord(XLogRecData *rdata,
 		/* advance global request to include new block(s) */
 		if (XLogCtl->LogwrtRqst.Write < EndPos)
 			XLogCtl->LogwrtRqst.Write = EndPos;
-		/* update local result copy while I have the chance */
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
+		/* update local result copy */
+		LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	}
 
 	/*
@@ -2170,13 +2169,13 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 			SpinLockAcquire(&XLogCtl->info_lck);
 			if (XLogCtl->LogwrtRqst.Write < OldPageRqstPtr)
 				XLogCtl->LogwrtRqst.Write = OldPageRqstPtr;
-			LogwrtResult = XLogCtl->LogwrtResult;
 			SpinLockRelease(&XLogCtl->info_lck);
 
 			/*
 			 * Now that we have an up-to-date LogwrtResult value, see if we
 			 * still need to write it or if someone else already did.
 			 */
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (LogwrtResult.Write < OldPageRqstPtr)
 			{
 				/*
@@ -2191,7 +2190,7 @@ AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic)
 
 				LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
 
-				LogwrtResult = XLogCtl->LogwrtResult;
+				LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 				if (LogwrtResult.Write >= OldPageRqstPtr)
 				{
 					/* OK, someone wrote it already */
@@ -2438,7 +2437,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	/*
 	 * Update local LogwrtResult (caller probably did this already, but...)
 	 */
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	/*
 	 * Since successive pages in the xlog cache are consecutively allocated,
@@ -2636,7 +2635,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 */
 	if (LogwrtResult.Flush < WriteRqst.Flush &&
 		LogwrtResult.Flush < LogwrtResult.Write)
-
 	{
 		/*
 		 * Could get here without iterating above loop, in which case we might
@@ -2675,8 +2673,10 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	 * code in a couple of places.
 	 */
 	{
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Write, LogwrtResult.Write);
+		pg_atomic_monotonic_advance_u64(&XLogCtl->LogwrtResult.Flush, LogwrtResult.Flush);
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
-		XLogCtl->LogwrtResult = LogwrtResult;
 		if (XLogCtl->LogwrtRqst.Write < LogwrtResult.Write)
 			XLogCtl->LogwrtRqst.Write = LogwrtResult.Write;
 		if (XLogCtl->LogwrtRqst.Flush < LogwrtResult.Flush)
@@ -2696,8 +2696,8 @@ XLogSetAsyncXactLSN(XLogRecPtr asyncXactLSN)
 	XLogRecPtr	WriteRqstPtr = asyncXactLSN;
 	bool		sleeping;
 
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	sleeping = XLogCtl->WalWriterSleeping;
 	if (XLogCtl->asyncXactLSN < asyncXactLSN)
 		XLogCtl->asyncXactLSN = asyncXactLSN;
@@ -2908,10 +2908,10 @@ XLogFlush(XLogRecPtr record)
 		SpinLockAcquire(&XLogCtl->info_lck);
 		if (WriteRqstPtr < XLogCtl->LogwrtRqst.Write)
 			WriteRqstPtr = XLogCtl->LogwrtRqst.Write;
-		LogwrtResult = XLogCtl->LogwrtResult;
 		SpinLockRelease(&XLogCtl->info_lck);
 
 		/* done already? */
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 			break;
 
@@ -2939,7 +2939,7 @@ XLogFlush(XLogRecPtr record)
 		}
 
 		/* Got the lock; recheck whether request is satisfied */
-		LogwrtResult = XLogCtl->LogwrtResult;
+		LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 		if (record <= LogwrtResult.Flush)
 		{
 			LWLockRelease(WALWriteLock);
@@ -3056,7 +3056,6 @@ XLogBackgroundFlush(void)
 
 	/* read LogwrtResult and update local state */
 	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
 	WriteRqst = XLogCtl->LogwrtRqst;
 	SpinLockRelease(&XLogCtl->info_lck);
 
@@ -3064,8 +3063,10 @@ XLogBackgroundFlush(void)
 	WriteRqst.Write -= WriteRqst.Write % XLOG_BLCKSZ;
 
 	/* if we have already flushed that far, consider async commit records */
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write <= LogwrtResult.Flush)
 	{
+		pg_memory_barrier();
 		SpinLockAcquire(&XLogCtl->info_lck);
 		WriteRqst.Write = XLogCtl->asyncXactLSN;
 		SpinLockRelease(&XLogCtl->info_lck);
@@ -3081,6 +3082,7 @@ XLogBackgroundFlush(void)
 	{
 		if (openLogFile >= 0)
 		{
+			LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 			if (!XLByteInPrevSeg(LogwrtResult.Write, openLogSegNo,
 								 wal_segment_size))
 			{
@@ -3140,7 +3142,8 @@ XLogBackgroundFlush(void)
 	/* now wait for any in-progress insertions to finish and get write lock */
 	WaitXLogInsertionsToFinish(WriteRqst.Write);
 	LWLockAcquire(WALWriteLock, LW_EXCLUSIVE);
-	LogwrtResult = XLogCtl->LogwrtResult;
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 	if (WriteRqst.Write > LogwrtResult.Write ||
 		WriteRqst.Flush > LogwrtResult.Flush)
 	{
@@ -3228,9 +3231,7 @@ XLogNeedsFlush(XLogRecPtr record)
 		return false;
 
 	/* read LogwrtResult and update local state */
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	/* check again */
 	if (record <= LogwrtResult.Flush)
@@ -7779,7 +7780,8 @@ StartupXLOG(void)
 
 	LogwrtResult.Write = LogwrtResult.Flush = EndOfLog;
 
-	XLogCtl->LogwrtResult = LogwrtResult;
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Write, EndOfLog);
+	pg_atomic_write_u64(&XLogCtl->LogwrtResult.Flush, EndOfLog);
 
 	XLogCtl->LogwrtRqst.Write = EndOfLog;
 	XLogCtl->LogwrtRqst.Flush = EndOfLog;
@@ -8498,9 +8500,7 @@ GetInsertRecPtr(void)
 XLogRecPtr
 GetFlushRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Flush = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Flush);
 
 	return LogwrtResult.Flush;
 }
@@ -11649,9 +11649,7 @@ GetXLogInsertRecPtr(void)
 XLogRecPtr
 GetXLogWriteRecPtr(void)
 {
-	SpinLockAcquire(&XLogCtl->info_lck);
-	LogwrtResult = XLogCtl->LogwrtResult;
-	SpinLockRelease(&XLogCtl->info_lck);
+	LogwrtResult.Write = pg_atomic_read_u64(&XLogCtl->LogwrtResult.Write);
 
 	return LogwrtResult.Write;
 }
-- 
2.20.1


--0OAP2g/MAC+5xKAE--





^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* Add test module for Table Access Method
@ 2023-06-03 23:42  Fabrízio de Royes Mello <[email protected]>
  0 siblings, 2 replies; 26+ messages in thread

From: Fabrízio de Royes Mello @ 2023-06-03 23:42 UTC (permalink / raw)
  To: pgsql-hackers

Hi all,

During the PGCon Unconference session about Table Access Method one missing
item pointed out is that currently we lack documentation and examples of
TAM.

So in order to improve things a bit in this area I'm proposing to add a
test module for Table Access Method similar what we already have for Index
Access Method.

This code is based on the "blackhole_am" implemented by Michael Paquier:
https://github.com/michaelpq/pg_plugins/tree/main/blackhole_am

Regards,

-- 
Fabrízio de Royes Mello


Attachments:

  [text/x-patch] 0001-Add-test-module-for-Table-Access-Method.patch (25.4K, ../../CAFcNs+pcU2ib=jvjNZNboD+M2tHO+vD77C_YZJ2rsGR0Tp35mg@mail.gmail.com/3-0001-Add-test-module-for-Table-Access-Method.patch)
  download | inline diff:
From 217b84f21ec1cdb0ede271d24b7a5863713db949 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fabr=C3=ADzio=20de=20Royes=20Mello?=
 <[email protected]>
Date: Sat, 3 Jun 2023 17:23:05 -0400
Subject: [PATCH] Add test module for Table Access Method

---
 src/test/modules/Makefile                     |   1 +
 src/test/modules/dummy_table_am/.gitignore    |   3 +
 src/test/modules/dummy_table_am/Makefile      |  20 +
 src/test/modules/dummy_table_am/README        |  12 +
 .../dummy_table_am/dummy_table_am--1.0.sql    |  14 +
 .../modules/dummy_table_am/dummy_table_am.c   | 519 ++++++++++++++++++
 .../dummy_table_am/dummy_table_am.control     |   5 +
 .../expected/dummy_table_am.out               | 127 +++++
 src/test/modules/dummy_table_am/meson.build   |  33 ++
 .../dummy_table_am/sql/dummy_table_am.sql     |  34 ++
 src/test/modules/meson.build                  |   1 +
 11 files changed, 769 insertions(+)
 create mode 100644 src/test/modules/dummy_table_am/.gitignore
 create mode 100644 src/test/modules/dummy_table_am/Makefile
 create mode 100644 src/test/modules/dummy_table_am/README
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am.c
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am.control
 create mode 100644 src/test/modules/dummy_table_am/expected/dummy_table_am.out
 create mode 100644 src/test/modules/dummy_table_am/meson.build
 create mode 100644 src/test/modules/dummy_table_am/sql/dummy_table_am.sql

diff --git a/src/test/modules/Makefile b/src/test/modules/Makefile
index 6331c976dc..ce982b0e46 100644
--- a/src/test/modules/Makefile
+++ b/src/test/modules/Makefile
@@ -10,6 +10,7 @@ SUBDIRS = \
 		  delay_execution \
 		  dummy_index_am \
 		  dummy_seclabel \
+		  dummy_table_am \
 		  libpq_pipeline \
 		  plsample \
 		  snapshot_too_old \
diff --git a/src/test/modules/dummy_table_am/.gitignore b/src/test/modules/dummy_table_am/.gitignore
new file mode 100644
index 0000000000..44d119cfcc
--- /dev/null
+++ b/src/test/modules/dummy_table_am/.gitignore
@@ -0,0 +1,3 @@
+# Generated subdirectories
+/log/
+/results/
diff --git a/src/test/modules/dummy_table_am/Makefile b/src/test/modules/dummy_table_am/Makefile
new file mode 100644
index 0000000000..9ea4a590c6
--- /dev/null
+++ b/src/test/modules/dummy_table_am/Makefile
@@ -0,0 +1,20 @@
+# src/test/modules/dummy_table_am/Makefile
+
+MODULES = dummy_table_am
+
+EXTENSION = dummy_table_am
+DATA = dummy_table_am--1.0.sql
+PGFILEDESC = "dummy_table_am - table access method template"
+
+REGRESS = dummy_table_am
+
+ifdef USE_PGXS
+PG_CONFIG = pg_config
+PGXS := $(shell $(PG_CONFIG) --pgxs)
+include $(PGXS)
+else
+subdir = src/test/modules/dummy_table_am
+top_builddir = ../../../..
+include $(top_builddir)/src/Makefile.global
+include $(top_srcdir)/contrib/contrib-global.mk
+endif
diff --git a/src/test/modules/dummy_table_am/README b/src/test/modules/dummy_table_am/README
new file mode 100644
index 0000000000..61510f02fa
--- /dev/null
+++ b/src/test/modules/dummy_table_am/README
@@ -0,0 +1,12 @@
+Dummy Index AM
+==============
+
+Dummy index AM is a module for testing any facility usable by an index
+access method, whose code is kept a maximum simple.
+
+This includes tests for all relation option types:
+- boolean
+- enum
+- integer
+- real
+- strings (with and without NULL as default)
diff --git a/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql b/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
new file mode 100644
index 0000000000..aa0fd82e61
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
@@ -0,0 +1,14 @@
+/* src/test/modules/dummy_table_am/dummy_table_am--1.0.sql */
+
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "CREATE EXTENSION dummy_table_am" to load this file. \quit
+
+CREATE FUNCTION dummy_table_am_handler(internal)
+RETURNS table_am_handler
+AS 'MODULE_PATHNAME'
+LANGUAGE C;
+
+-- Access method
+CREATE ACCESS METHOD dummy_table_am TYPE TABLE HANDLER dummy_table_am_handler;
+COMMENT ON ACCESS METHOD dummy_table_am IS 'dummy table access method';
+
diff --git a/src/test/modules/dummy_table_am/dummy_table_am.c b/src/test/modules/dummy_table_am/dummy_table_am.c
new file mode 100644
index 0000000000..b299fe9c65
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am.c
@@ -0,0 +1,519 @@
+/*-------------------------------------------------------------------------
+ *
+ * dummy_table_am.c
+ *		Index AM template main file.
+ *
+ * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
+ *
+ * IDENTIFICATION
+ *	  src/test/modules/dummy_table_am/dummy_table_am.c
+ *
+ *-------------------------------------------------------------------------
+ */
+#include "postgres.h"
+
+#include "access/heapam.h"
+#include "access/tableam.h"
+#include "catalog/index.h"
+#include "commands/vacuum.h"
+
+PG_MODULE_MAGIC;
+
+/* Handler for table AM */
+PG_FUNCTION_INFO_V1(dummy_table_am_handler);
+
+/* Base structures for scans */
+typedef struct DummyScanDescData
+{
+	TableScanDescData rs_base;
+} DummyScanDescData;
+
+typedef struct DummyScanDescData *DummyScanDesc;
+
+/*
+ * Slot related callbacks for Dummy Table Access Method
+ */
+static const TupleTableSlotOps *
+dummy_table_am_slot_callbacks(Relation relation)
+{
+	elog(INFO, "%s", __func__);
+	return &TTSOpsMinimalTuple;
+}
+
+/*
+ * Table Scan Callbacks for dummy_table_am AM
+ */
+static TableScanDesc
+dummy_table_am_scan_begin(Relation relation, Snapshot snapshot,
+					 int nkeys, ScanKey key,
+					 ParallelTableScanDesc parallel_scan,
+					 uint32 flags)
+{
+	DummyScanDesc scan;
+
+	scan = (DummyScanDesc) palloc(sizeof(DummyScanDescData));
+
+	scan->rs_base.rs_rd = relation;
+	scan->rs_base.rs_snapshot = snapshot;
+	scan->rs_base.rs_nkeys = nkeys;
+	scan->rs_base.rs_flags = flags;
+	scan->rs_base.rs_parallel = parallel_scan;
+
+	elog(INFO, "%s", __func__);
+
+	return (TableScanDesc) scan;
+}
+
+static void
+dummy_table_am_scan_end(TableScanDesc sscan)
+{
+	DummyScanDesc scan = (DummyScanDesc) sscan;
+
+	elog(INFO, "%s", __func__);
+
+	pfree(scan);
+}
+
+static void
+dummy_table_am_scan_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
+					  bool allow_strat, bool allow_sync, bool allow_pagemode)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_scan_getnextslot(TableScanDesc sscan, ScanDirection direction,
+						   TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+/*
+ * Index Scan Callbacks for dummy_table_am AM
+ */
+static IndexFetchTableData *
+dummy_table_am_index_fetch_begin(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+	return NULL;
+}
+
+static void
+dummy_table_am_index_fetch_reset(IndexFetchTableData *scan)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_index_fetch_end(IndexFetchTableData *scan)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_index_fetch_tuple(struct IndexFetchTableData *scan,
+							ItemPointer tid,
+							Snapshot snapshot,
+							TupleTableSlot *slot,
+							bool *call_again, bool *all_dead)
+{
+	elog(INFO, "%s", __func__);
+	return 0;
+}
+
+
+/* ------------------------------------------------------------------------
+ * Callbacks for non-modifying operations on individual tuples for
+ * dummy_table_am AM.
+ * ------------------------------------------------------------------------
+ */
+
+static bool
+dummy_table_am_fetch_row_version(Relation relation,
+							ItemPointer tid,
+							Snapshot snapshot,
+							TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static void
+dummy_table_am_get_latest_tid(TableScanDesc sscan,
+						 ItemPointer tid)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static bool
+dummy_table_am_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
+								   Snapshot snapshot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static TransactionId
+dummy_table_am_index_delete_tuples(Relation rel,
+							  TM_IndexDeleteOp *delstate)
+{
+	elog(INFO, "%s", __func__);
+	return InvalidTransactionId;
+}
+
+/* ----------------------------------------------------------------------------
+ *  Functions for manipulations of physical tuples for dummy_table_am AM.
+ * ----------------------------------------------------------------------------
+ */
+
+static void
+dummy_table_am_tuple_insert(Relation relation, TupleTableSlot *slot,
+					   CommandId cid, int options, BulkInsertState bistate)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_tuple_insert_speculative(Relation relation, TupleTableSlot *slot,
+								   CommandId cid, int options,
+								   BulkInsertState bistate,
+								   uint32 specToken)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_tuple_complete_speculative(Relation relation, TupleTableSlot *slot,
+									 uint32 spekToken, bool succeeded)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_multi_insert(Relation relation, TupleTableSlot **slots,
+					   int ntuples, CommandId cid, int options,
+					   BulkInsertState bistate)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static TM_Result
+dummy_table_am_tuple_delete(Relation relation, ItemPointer tid, CommandId cid,
+					   Snapshot snapshot, Snapshot crosscheck, bool wait,
+					   TM_FailureData *tmfd, bool changingPart)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+
+static TM_Result
+dummy_table_am_tuple_update(Relation relation, ItemPointer otid,
+					   TupleTableSlot *slot, CommandId cid,
+					   Snapshot snapshot, Snapshot crosscheck,
+					   bool wait, TM_FailureData *tmfd,
+					   LockTupleMode *lockmode,
+					   TU_UpdateIndexes *update_indexes)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+static TM_Result
+dummy_table_am_tuple_lock(Relation relation, ItemPointer tid, Snapshot snapshot,
+					 TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
+					 LockWaitPolicy wait_policy, uint8 flags,
+					 TM_FailureData *tmfd)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+static void
+dummy_table_am_finish_bulk_insert(Relation relation, int options)
+{
+	elog(INFO, "%s", __func__);
+}
+
+
+/* ------------------------------------------------------------------------
+ * DDL related callbacks for dummy_table_am AM.
+ * ------------------------------------------------------------------------
+ */
+
+static void
+dummy_table_am_relation_set_new_filelocator(Relation rel,
+									   const RelFileLocator *newrnode,
+									   char persistence,
+									   TransactionId *freezeXid,
+									   MultiXactId *minmulti)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_relation_nontransactional_truncate(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_copy_data(Relation rel, const RelFileLocator *newrnode)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_copy_for_cluster(Relation OldTable, Relation NewTable,
+						   Relation OldIndex, bool use_sort,
+						   TransactionId OldestXmin,
+						   TransactionId *xid_cutoff,
+						   MultiXactId *multi_cutoff,
+						   double *num_tuples,
+						   double *tups_vacuumed,
+						   double *tups_recently_dead)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_vacuum(Relation onerel, VacuumParams *params,
+				 BufferAccessStrategy bstrategy)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_scan_analyze_next_block(TableScanDesc scan, BlockNumber blockno,
+								  BufferAccessStrategy bstrategy)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to analyze next block */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
+								  double *liverows, double *deadrows,
+								  TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to analyze next tuple */
+	return false;
+}
+
+static double
+dummy_table_am_index_build_range_scan(Relation tableRelation,
+								 Relation indexRelation,
+								 IndexInfo *indexInfo,
+								 bool allow_sync,
+								 bool anyvisible,
+								 bool progress,
+								 BlockNumber start_blockno,
+								 BlockNumber numblocks,
+								 IndexBuildCallback callback,
+								 void *callback_state,
+								 TableScanDesc scan)
+{
+	elog(ERROR, "%s", __func__);
+
+	/* no data, so no tuples */
+	return 0;
+}
+
+static void
+dummy_table_am_index_validate_scan(Relation tableRelation,
+							  Relation indexRelation,
+							  IndexInfo *indexInfo,
+							  Snapshot snapshot,
+							  ValidateIndexState *state)
+{
+	elog(INFO, "%s", __func__);
+}
+
+
+/* ------------------------------------------------------------------------
+ * Miscellaneous callbacks for the dummy_table_am AM
+ * ------------------------------------------------------------------------
+ */
+
+static uint64
+dummy_table_am_relation_size(Relation rel, ForkNumber forkNumber)
+{
+	elog(INFO, "%s", __func__);
+
+	/* there is nothing */
+	return 0;
+}
+
+/*
+ * Check to see whether the table needs a TOAST table.
+ */
+static bool
+dummy_table_am_relation_needs_toast_table(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no toast table needed */
+	return false;
+}
+
+
+/* ------------------------------------------------------------------------
+ * Planner related callbacks for the dummy_table_am AM
+ * ------------------------------------------------------------------------
+ */
+
+static void
+dummy_table_am_estimate_rel_size(Relation rel, int32 *attr_widths,
+							BlockNumber *pages, double *tuples,
+							double *allvisfrac)
+{
+	/* no data available */
+	if (attr_widths)
+		*attr_widths = 0;
+	if (pages)
+		*pages = 0;
+	if (tuples)
+		*tuples = 0;
+	if (allvisfrac)
+		*allvisfrac = 0;
+
+	elog(INFO, "%s", __func__);
+}
+
+
+/* ------------------------------------------------------------------------
+ * Executor related callbacks for the dummy_table_am AM
+ * ------------------------------------------------------------------------
+ */
+
+static bool
+dummy_table_am_scan_bitmap_next_block(TableScanDesc scan,
+								 TBMIterateResult *tbmres)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next block */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_bitmap_next_tuple(TableScanDesc scan,
+								 TBMIterateResult *tbmres,
+								 TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next tuple */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_sample_next_block(TableScanDesc scan,
+								 SampleScanState *scanstate)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next block for sampling */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_sample_next_tuple(TableScanDesc scan,
+								 SampleScanState *scanstate,
+								 TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next tuple for sampling */
+	return false;
+}
+
+/* ------------------------------------------------------------------------
+ * Definition of the dummy_table_am table access method.
+ * ------------------------------------------------------------------------
+ */
+
+static const TableAmRoutine dummy_table_am_methods = {
+	.type = T_TableAmRoutine,
+
+	.slot_callbacks = dummy_table_am_slot_callbacks,
+
+	.scan_begin = dummy_table_am_scan_begin,
+	.scan_end = dummy_table_am_scan_end,
+	.scan_rescan = dummy_table_am_scan_rescan,
+	.scan_getnextslot = dummy_table_am_scan_getnextslot,
+
+	/* these are common helper functions */
+	.parallelscan_estimate = table_block_parallelscan_estimate,
+	.parallelscan_initialize = table_block_parallelscan_initialize,
+	.parallelscan_reinitialize = table_block_parallelscan_reinitialize,
+
+	.index_fetch_begin = dummy_table_am_index_fetch_begin,
+	.index_fetch_reset = dummy_table_am_index_fetch_reset,
+	.index_fetch_end = dummy_table_am_index_fetch_end,
+	.index_fetch_tuple = dummy_table_am_index_fetch_tuple,
+
+	.tuple_insert = dummy_table_am_tuple_insert,
+	.tuple_insert_speculative = dummy_table_am_tuple_insert_speculative,
+	.tuple_complete_speculative = dummy_table_am_tuple_complete_speculative,
+	.multi_insert = dummy_table_am_multi_insert,
+	.tuple_delete = dummy_table_am_tuple_delete,
+	.tuple_update = dummy_table_am_tuple_update,
+	.tuple_lock = dummy_table_am_tuple_lock,
+	.finish_bulk_insert = dummy_table_am_finish_bulk_insert,
+
+	.tuple_fetch_row_version = dummy_table_am_fetch_row_version,
+	.tuple_get_latest_tid = dummy_table_am_get_latest_tid,
+	.tuple_tid_valid = dummy_table_am_tuple_tid_valid,
+	.tuple_satisfies_snapshot = dummy_table_am_tuple_satisfies_snapshot,
+	.index_delete_tuples = dummy_table_am_index_delete_tuples,
+
+	.relation_set_new_filelocator = dummy_table_am_relation_set_new_filelocator,
+	.relation_nontransactional_truncate = dummy_table_am_relation_nontransactional_truncate,
+	.relation_copy_data = dummy_table_am_copy_data,
+	.relation_copy_for_cluster = dummy_table_am_copy_for_cluster,
+	.relation_vacuum = dummy_table_am_vacuum,
+	.scan_analyze_next_block = dummy_table_am_scan_analyze_next_block,
+	.scan_analyze_next_tuple = dummy_table_am_scan_analyze_next_tuple,
+	.index_build_range_scan = dummy_table_am_index_build_range_scan,
+	.index_validate_scan = dummy_table_am_index_validate_scan,
+
+	.relation_size = dummy_table_am_relation_size,
+	.relation_needs_toast_table = dummy_table_am_relation_needs_toast_table,
+
+	.relation_estimate_size = dummy_table_am_estimate_rel_size,
+
+	.scan_bitmap_next_block = dummy_table_am_scan_bitmap_next_block,
+	.scan_bitmap_next_tuple = dummy_table_am_scan_bitmap_next_tuple,
+	.scan_sample_next_block = dummy_table_am_scan_sample_next_block,
+	.scan_sample_next_tuple = dummy_table_am_scan_sample_next_tuple
+};
+
+/*
+ * Table AM handler function: returns TableAmRoutine with access method
+ * parameters and callbacks.
+ */
+Datum
+dummy_table_am_handler(PG_FUNCTION_ARGS)
+{
+	PG_RETURN_POINTER(&dummy_table_am_methods);
+}
diff --git a/src/test/modules/dummy_table_am/dummy_table_am.control b/src/test/modules/dummy_table_am/dummy_table_am.control
new file mode 100644
index 0000000000..08f2f868d4
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am.control
@@ -0,0 +1,5 @@
+# dummy_table_am extension
+comment = 'dummy_table_am - table access method template'
+default_version = '1.0'
+module_pathname = '$libdir/dummy_table_am'
+relocatable = true
diff --git a/src/test/modules/dummy_table_am/expected/dummy_table_am.out b/src/test/modules/dummy_table_am/expected/dummy_table_am.out
new file mode 100644
index 0000000000..758dc1b9eb
--- /dev/null
+++ b/src/test/modules/dummy_table_am/expected/dummy_table_am.out
@@ -0,0 +1,127 @@
+-- Tests for dummy table access method
+CREATE EXTENSION dummy_table_am;
+CREATE TABLE dummy_table (a int, b text) USING dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+-- Will error out
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INFO:  dummy_table_am_estimate_rel_size
+ERROR:  dummy_table_am_index_build_range_scan
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+INSERT INTO dummy_table VALUES (1, 'dummy');
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+UPDATE dummy_table SET a = 0 WHERE a = 1;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+DELETE FROM dummy_table WHERE a = 1;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+TRUNCATE dummy_table;
+INFO:  dummy_table_am_relation_set_new_filelocator
+VACUUM dummy_table;
+INFO:  dummy_table_am_vacuum
+ANALYZE dummy_table;
+INFO:  dummy_table_am_relation_size
+INFO:  dummy_table_am_relation_size
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_end
+COPY dummy_table TO STDOUT;
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+COPY dummy_table (a, b) FROM STDIN;
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_multi_insert
+INFO:  dummy_table_am_finish_bulk_insert
+-- ALTER TABLE SET ACCESS METHOD
+ALTER TABLE dummy_table SET ACCESS METHOD heap;
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INSERT INTO dummy_table VALUES (1, 'heap');
+SELECT * FROM dummy_table;
+ a |  b   
+---+------
+ 1 | heap
+(1 row)
+
+-- Will error out, the index must be dropped
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+INFO:  dummy_table_am_finish_bulk_insert
+INFO:  dummy_table_am_estimate_rel_size
+ERROR:  dummy_table_am_index_build_range_scan
+DROP INDEX dummy_table_idx;
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+INFO:  dummy_table_am_finish_bulk_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+-- Clean up
+DROP TABLE dummy_table;
diff --git a/src/test/modules/dummy_table_am/meson.build b/src/test/modules/dummy_table_am/meson.build
new file mode 100644
index 0000000000..93f9108f29
--- /dev/null
+++ b/src/test/modules/dummy_table_am/meson.build
@@ -0,0 +1,33 @@
+# Copyright (c) 2022-2023, PostgreSQL Global Development Group
+
+dummy_table_am_sources = files(
+  'dummy_table_am.c',
+)
+
+if host_system == 'windows'
+  dummy_table_am_sources += rc_lib_gen.process(win32ver_rc, extra_args: [
+    '--NAME', 'dummy_table_am',
+    '--FILEDESC', 'dummy_table_am - table access method template',])
+endif
+
+dummy_table_am = shared_module('dummy_table_am',
+  dummy_table_am_sources,
+  kwargs: pg_test_mod_args,
+)
+test_install_libs += dummy_table_am
+
+test_install_data += files(
+  'dummy_table_am.control',
+  'dummy_table_am--1.0.sql',
+)
+
+tests += {
+  'name': 'dummy_table_am',
+  'sd': meson.current_source_dir(),
+  'bd': meson.current_build_dir(),
+  'regress': {
+    'sql': [
+      'dummy_table_am',
+    ],
+  },
+}
diff --git a/src/test/modules/dummy_table_am/sql/dummy_table_am.sql b/src/test/modules/dummy_table_am/sql/dummy_table_am.sql
new file mode 100644
index 0000000000..ae693683ab
--- /dev/null
+++ b/src/test/modules/dummy_table_am/sql/dummy_table_am.sql
@@ -0,0 +1,34 @@
+-- Tests for dummy table access method
+CREATE EXTENSION dummy_table_am;
+CREATE TABLE dummy_table (a int, b text) USING dummy_table_am;
+-- Will error out
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+SELECT * FROM dummy_table;
+INSERT INTO dummy_table VALUES (1, 'dummy');
+SELECT * FROM dummy_table;
+UPDATE dummy_table SET a = 0 WHERE a = 1;
+SELECT * FROM dummy_table;
+DELETE FROM dummy_table WHERE a = 1;
+SELECT * FROM dummy_table;
+TRUNCATE dummy_table;
+VACUUM dummy_table;
+ANALYZE dummy_table;
+COPY dummy_table TO STDOUT;
+COPY dummy_table (a, b) FROM STDIN;
+1	dummy
+\.
+
+-- ALTER TABLE SET ACCESS METHOD
+ALTER TABLE dummy_table SET ACCESS METHOD heap;
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INSERT INTO dummy_table VALUES (1, 'heap');
+SELECT * FROM dummy_table;
+
+-- Will error out, the index must be dropped
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+DROP INDEX dummy_table_idx;
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+SELECT * FROM dummy_table;
+
+-- Clean up
+DROP TABLE dummy_table;
diff --git a/src/test/modules/meson.build b/src/test/modules/meson.build
index 17d369e378..84460f27b4 100644
--- a/src/test/modules/meson.build
+++ b/src/test/modules/meson.build
@@ -5,6 +5,7 @@ subdir('commit_ts')
 subdir('delay_execution')
 subdir('dummy_index_am')
 subdir('dummy_seclabel')
+subdir('dummy_table_am')
 subdir('ldap_password_func')
 subdir('libpq_pipeline')
 subdir('plsample')
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* Re: Add test module for Table Access Method
@ 2023-06-05 16:24  Fabrízio de Royes Mello <[email protected]>
  parent: Fabrízio de Royes Mello <[email protected]>
  1 sibling, 0 replies; 26+ messages in thread

From: Fabrízio de Royes Mello @ 2023-06-05 16:24 UTC (permalink / raw)
  To: pgsql-hackers

On Sat, Jun 3, 2023 at 7:42 PM Fabrízio de Royes Mello <
[email protected]> wrote:
>
>
> Hi all,
>
> During the PGCon Unconference session about Table Access Method one
missing item pointed out is that currently we lack documentation and
examples of TAM.
>
> So in order to improve things a bit in this area I'm proposing to add a
test module for Table Access Method similar what we already have for Index
Access Method.
>
> This code is based on the "blackhole_am" implemented by Michael Paquier:
https://github.com/michaelpq/pg_plugins/tree/main/blackhole_am
>

Just added some more tests, ran pgindent and also organized a bit some
comments and README.txt.

Regards,

-- 
Fabrízio de Royes Mello


Attachments:

  [text/x-patch] v2-0001-Add-test-module-for-Table-Access-Method.patch (26.9K, ../../CAFcNs+qPD5U3vUzHxr2DGPgV_E14OpSj-ypoLk84QKN=Jux1=A@mail.gmail.com/3-v2-0001-Add-test-module-for-Table-Access-Method.patch)
  download | inline diff:
From 31d6bf00fbee6c229382db0760ba602e4d41c917 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fabr=C3=ADzio=20de=20Royes=20Mello?=
 <[email protected]>
Date: Sat, 3 Jun 2023 17:23:05 -0400
Subject: [PATCH v2] Add test module for Table Access Method

---
 src/test/modules/Makefile                     |   1 +
 src/test/modules/dummy_table_am/.gitignore    |   3 +
 src/test/modules/dummy_table_am/Makefile      |  20 +
 src/test/modules/dummy_table_am/README        |   5 +
 .../dummy_table_am/dummy_table_am--1.0.sql    |  14 +
 .../modules/dummy_table_am/dummy_table_am.c   | 500 ++++++++++++++++++
 .../dummy_table_am/dummy_table_am.control     |   5 +
 .../expected/dummy_table_am.out               | 207 ++++++++
 src/test/modules/dummy_table_am/meson.build   |  33 ++
 .../dummy_table_am/sql/dummy_table_am.sql     |  55 ++
 src/test/modules/meson.build                  |   1 +
 11 files changed, 844 insertions(+)
 create mode 100644 src/test/modules/dummy_table_am/.gitignore
 create mode 100644 src/test/modules/dummy_table_am/Makefile
 create mode 100644 src/test/modules/dummy_table_am/README
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am.c
 create mode 100644 src/test/modules/dummy_table_am/dummy_table_am.control
 create mode 100644 src/test/modules/dummy_table_am/expected/dummy_table_am.out
 create mode 100644 src/test/modules/dummy_table_am/meson.build
 create mode 100644 src/test/modules/dummy_table_am/sql/dummy_table_am.sql

diff --git a/src/test/modules/Makefile b/src/test/modules/Makefile
index 6331c976dc..ce982b0e46 100644
--- a/src/test/modules/Makefile
+++ b/src/test/modules/Makefile
@@ -10,6 +10,7 @@ SUBDIRS = \
 		  delay_execution \
 		  dummy_index_am \
 		  dummy_seclabel \
+		  dummy_table_am \
 		  libpq_pipeline \
 		  plsample \
 		  snapshot_too_old \
diff --git a/src/test/modules/dummy_table_am/.gitignore b/src/test/modules/dummy_table_am/.gitignore
new file mode 100644
index 0000000000..44d119cfcc
--- /dev/null
+++ b/src/test/modules/dummy_table_am/.gitignore
@@ -0,0 +1,3 @@
+# Generated subdirectories
+/log/
+/results/
diff --git a/src/test/modules/dummy_table_am/Makefile b/src/test/modules/dummy_table_am/Makefile
new file mode 100644
index 0000000000..9ea4a590c6
--- /dev/null
+++ b/src/test/modules/dummy_table_am/Makefile
@@ -0,0 +1,20 @@
+# src/test/modules/dummy_table_am/Makefile
+
+MODULES = dummy_table_am
+
+EXTENSION = dummy_table_am
+DATA = dummy_table_am--1.0.sql
+PGFILEDESC = "dummy_table_am - table access method template"
+
+REGRESS = dummy_table_am
+
+ifdef USE_PGXS
+PG_CONFIG = pg_config
+PGXS := $(shell $(PG_CONFIG) --pgxs)
+include $(PGXS)
+else
+subdir = src/test/modules/dummy_table_am
+top_builddir = ../../../..
+include $(top_builddir)/src/Makefile.global
+include $(top_srcdir)/contrib/contrib-global.mk
+endif
diff --git a/src/test/modules/dummy_table_am/README b/src/test/modules/dummy_table_am/README
new file mode 100644
index 0000000000..35211554b0
--- /dev/null
+++ b/src/test/modules/dummy_table_am/README
@@ -0,0 +1,5 @@
+Dummy Table AM
+==============
+
+Dummy table AM is a module for testing any facility usable by an table
+access method, whose code is kept a maximum simple.
diff --git a/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql b/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
new file mode 100644
index 0000000000..aa0fd82e61
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am--1.0.sql
@@ -0,0 +1,14 @@
+/* src/test/modules/dummy_table_am/dummy_table_am--1.0.sql */
+
+-- complain if script is sourced in psql, rather than via CREATE EXTENSION
+\echo Use "CREATE EXTENSION dummy_table_am" to load this file. \quit
+
+CREATE FUNCTION dummy_table_am_handler(internal)
+RETURNS table_am_handler
+AS 'MODULE_PATHNAME'
+LANGUAGE C;
+
+-- Access method
+CREATE ACCESS METHOD dummy_table_am TYPE TABLE HANDLER dummy_table_am_handler;
+COMMENT ON ACCESS METHOD dummy_table_am IS 'dummy table access method';
+
diff --git a/src/test/modules/dummy_table_am/dummy_table_am.c b/src/test/modules/dummy_table_am/dummy_table_am.c
new file mode 100644
index 0000000000..132f7b18cd
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am.c
@@ -0,0 +1,500 @@
+/*-------------------------------------------------------------------------
+ *
+ * dummy_table_am.c
+ *		Index AM template main file.
+ *
+ * Portions Copyright (c) 1996-2023, PostgreSQL Global Development Group
+ * Portions Copyright (c) 1994, Regents of the University of California
+ *
+ * IDENTIFICATION
+ *	  src/test/modules/dummy_table_am/dummy_table_am.c
+ *
+ *-------------------------------------------------------------------------
+ */
+#include "postgres.h"
+
+#include "access/heapam.h"
+#include "access/tableam.h"
+#include "catalog/index.h"
+#include "commands/vacuum.h"
+
+PG_MODULE_MAGIC;
+
+/* Handler for table AM */
+PG_FUNCTION_INFO_V1(dummy_table_am_handler);
+
+/* Base structures for scans */
+typedef struct DummyScanDescData
+{
+	TableScanDescData rs_base;
+}			DummyScanDescData;
+
+typedef struct DummyScanDescData *DummyScanDesc;
+
+/*
+ * Slot related callbacks for Dummy Table Access Method
+ */
+static const TupleTableSlotOps *
+dummy_table_am_slot_callbacks(Relation relation)
+{
+	elog(INFO, "%s", __func__);
+	return &TTSOpsMinimalTuple;
+}
+
+/*
+ * Table Scan Callbacks for dummy_table_am AM
+ */
+static TableScanDesc
+dummy_table_am_scan_begin(Relation relation, Snapshot snapshot,
+						  int nkeys, ScanKey key,
+						  ParallelTableScanDesc parallel_scan,
+						  uint32 flags)
+{
+	DummyScanDesc scan;
+
+	scan = (DummyScanDesc) palloc(sizeof(DummyScanDescData));
+
+	scan->rs_base.rs_rd = relation;
+	scan->rs_base.rs_snapshot = snapshot;
+	scan->rs_base.rs_nkeys = nkeys;
+	scan->rs_base.rs_flags = flags;
+	scan->rs_base.rs_parallel = parallel_scan;
+
+	elog(INFO, "%s", __func__);
+
+	return (TableScanDesc) scan;
+}
+
+static void
+dummy_table_am_scan_end(TableScanDesc sscan)
+{
+	DummyScanDesc scan = (DummyScanDesc) sscan;
+
+	elog(INFO, "%s", __func__);
+
+	pfree(scan);
+}
+
+static void
+dummy_table_am_scan_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
+						   bool allow_strat, bool allow_sync, bool allow_pagemode)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_scan_getnextslot(TableScanDesc sscan, ScanDirection direction,
+								TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+/*
+ * Index Scan Callbacks for dummy_table_am AM
+ */
+static IndexFetchTableData *
+dummy_table_am_index_fetch_begin(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+	return NULL;
+}
+
+static void
+dummy_table_am_index_fetch_reset(IndexFetchTableData *scan)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_index_fetch_end(IndexFetchTableData *scan)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_index_fetch_tuple(struct IndexFetchTableData *scan,
+								 ItemPointer tid,
+								 Snapshot snapshot,
+								 TupleTableSlot *slot,
+								 bool *call_again, bool *all_dead)
+{
+	elog(INFO, "%s", __func__);
+	return 0;
+}
+
+/*
+ * Callbacks for non-modifying operations on individual tuples for
+ * dummy_table_am AM.
+ */
+static bool
+dummy_table_am_fetch_row_version(Relation relation,
+								 ItemPointer tid,
+								 Snapshot snapshot,
+								 TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static void
+dummy_table_am_get_latest_tid(TableScanDesc sscan,
+							  ItemPointer tid)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_tuple_tid_valid(TableScanDesc scan, ItemPointer tid)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static bool
+dummy_table_am_tuple_satisfies_snapshot(Relation rel, TupleTableSlot *slot,
+										Snapshot snapshot)
+{
+	elog(INFO, "%s", __func__);
+	return false;
+}
+
+static TransactionId
+dummy_table_am_index_delete_tuples(Relation rel,
+								   TM_IndexDeleteOp *delstate)
+{
+	elog(INFO, "%s", __func__);
+	return InvalidTransactionId;
+}
+
+/*
+ * Functions for manipulations of physical tuples for dummy_table_am AM.
+ */
+static void
+dummy_table_am_tuple_insert(Relation relation, TupleTableSlot *slot,
+							CommandId cid, int options, BulkInsertState bistate)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_tuple_insert_speculative(Relation relation, TupleTableSlot *slot,
+										CommandId cid, int options,
+										BulkInsertState bistate,
+										uint32 specToken)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_tuple_complete_speculative(Relation relation, TupleTableSlot *slot,
+										  uint32 spekToken, bool succeeded)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_multi_insert(Relation relation, TupleTableSlot **slots,
+							int ntuples, CommandId cid, int options,
+							BulkInsertState bistate)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static TM_Result
+dummy_table_am_tuple_delete(Relation relation, ItemPointer tid, CommandId cid,
+							Snapshot snapshot, Snapshot crosscheck, bool wait,
+							TM_FailureData *tmfd, bool changingPart)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+static TM_Result
+dummy_table_am_tuple_update(Relation relation, ItemPointer otid,
+							TupleTableSlot *slot, CommandId cid,
+							Snapshot snapshot, Snapshot crosscheck,
+							bool wait, TM_FailureData *tmfd,
+							LockTupleMode *lockmode,
+							TU_UpdateIndexes *update_indexes)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+static TM_Result
+dummy_table_am_tuple_lock(Relation relation, ItemPointer tid, Snapshot snapshot,
+						  TupleTableSlot *slot, CommandId cid, LockTupleMode mode,
+						  LockWaitPolicy wait_policy, uint8 flags,
+						  TM_FailureData *tmfd)
+{
+	elog(INFO, "%s", __func__);
+
+	/* nothing to do, so it is always OK */
+	return TM_Ok;
+}
+
+static void
+dummy_table_am_finish_bulk_insert(Relation relation, int options)
+{
+	elog(INFO, "%s", __func__);
+}
+
+/*
+ * DDL related callbacks for dummy_table_am AM.
+ */
+static void
+dummy_table_am_relation_set_new_filelocator(Relation rel,
+											const RelFileLocator *newrnode,
+											char persistence,
+											TransactionId *freezeXid,
+											MultiXactId *minmulti)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_relation_nontransactional_truncate(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_copy_data(Relation rel, const RelFileLocator *newrnode)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_copy_for_cluster(Relation OldTable, Relation NewTable,
+								Relation OldIndex, bool use_sort,
+								TransactionId OldestXmin,
+								TransactionId *xid_cutoff,
+								MultiXactId *multi_cutoff,
+								double *num_tuples,
+								double *tups_vacuumed,
+								double *tups_recently_dead)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static void
+dummy_table_am_vacuum(Relation onerel, VacuumParams *params,
+					  BufferAccessStrategy bstrategy)
+{
+	elog(INFO, "%s", __func__);
+}
+
+static bool
+dummy_table_am_scan_analyze_next_block(TableScanDesc scan, BlockNumber blockno,
+									   BufferAccessStrategy bstrategy)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to analyze next block */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
+									   double *liverows, double *deadrows,
+									   TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to analyze next tuple */
+	return false;
+}
+
+static double
+dummy_table_am_index_build_range_scan(Relation tableRelation,
+									  Relation indexRelation,
+									  IndexInfo *indexInfo,
+									  bool allow_sync,
+									  bool anyvisible,
+									  bool progress,
+									  BlockNumber start_blockno,
+									  BlockNumber numblocks,
+									  IndexBuildCallback callback,
+									  void *callback_state,
+									  TableScanDesc scan)
+{
+	elog(ERROR, "%s", __func__);
+
+	/* no data, so no tuples */
+	return 0;
+}
+
+static void
+dummy_table_am_index_validate_scan(Relation tableRelation,
+								   Relation indexRelation,
+								   IndexInfo *indexInfo,
+								   Snapshot snapshot,
+								   ValidateIndexState *state)
+{
+	elog(INFO, "%s", __func__);
+}
+
+/*
+ * Miscellaneous callbacks for the dummy_table_am AM
+ */
+static uint64
+dummy_table_am_relation_size(Relation rel, ForkNumber forkNumber)
+{
+	elog(INFO, "%s", __func__);
+
+	/* there is nothing */
+	return 0;
+}
+
+/*
+ * Check to see whether the table needs a TOAST table.
+ */
+static bool
+dummy_table_am_relation_needs_toast_table(Relation rel)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no toast table needed */
+	return false;
+}
+
+/*
+ * Planner related callbacks for the dummy_table_am AM
+ */
+static void
+dummy_table_am_estimate_rel_size(Relation rel, int32 *attr_widths,
+								 BlockNumber *pages, double *tuples,
+								 double *allvisfrac)
+{
+	/* no data available */
+	if (attr_widths)
+		*attr_widths = 0;
+	if (pages)
+		*pages = 0;
+	if (tuples)
+		*tuples = 0;
+	if (allvisfrac)
+		*allvisfrac = 0;
+
+	elog(INFO, "%s", __func__);
+}
+
+
+/*
+ * Executor related callbacks for the dummy_table_am AM
+ */
+static bool
+dummy_table_am_scan_bitmap_next_block(TableScanDesc scan,
+									  TBMIterateResult *tbmres)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next block */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_bitmap_next_tuple(TableScanDesc scan,
+									  TBMIterateResult *tbmres,
+									  TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next tuple */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_sample_next_block(TableScanDesc scan,
+									  SampleScanState *scanstate)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next block for sampling */
+	return false;
+}
+
+static bool
+dummy_table_am_scan_sample_next_tuple(TableScanDesc scan,
+									  SampleScanState *scanstate,
+									  TupleTableSlot *slot)
+{
+	elog(INFO, "%s", __func__);
+
+	/* no data, so no point to scan next tuple for sampling */
+	return false;
+}
+
+/*
+ * Definition of the dummy_table_am table access method.
+ */
+static const TableAmRoutine dummy_table_am_methods = {
+	.type = T_TableAmRoutine,
+
+	.slot_callbacks = dummy_table_am_slot_callbacks,
+
+	.scan_begin = dummy_table_am_scan_begin,
+	.scan_end = dummy_table_am_scan_end,
+	.scan_rescan = dummy_table_am_scan_rescan,
+	.scan_getnextslot = dummy_table_am_scan_getnextslot,
+
+	/* these are common helper functions */
+	.parallelscan_estimate = table_block_parallelscan_estimate,
+	.parallelscan_initialize = table_block_parallelscan_initialize,
+	.parallelscan_reinitialize = table_block_parallelscan_reinitialize,
+
+	.index_fetch_begin = dummy_table_am_index_fetch_begin,
+	.index_fetch_reset = dummy_table_am_index_fetch_reset,
+	.index_fetch_end = dummy_table_am_index_fetch_end,
+	.index_fetch_tuple = dummy_table_am_index_fetch_tuple,
+
+	.tuple_insert = dummy_table_am_tuple_insert,
+	.tuple_insert_speculative = dummy_table_am_tuple_insert_speculative,
+	.tuple_complete_speculative = dummy_table_am_tuple_complete_speculative,
+	.multi_insert = dummy_table_am_multi_insert,
+	.tuple_delete = dummy_table_am_tuple_delete,
+	.tuple_update = dummy_table_am_tuple_update,
+	.tuple_lock = dummy_table_am_tuple_lock,
+	.finish_bulk_insert = dummy_table_am_finish_bulk_insert,
+
+	.tuple_fetch_row_version = dummy_table_am_fetch_row_version,
+	.tuple_get_latest_tid = dummy_table_am_get_latest_tid,
+	.tuple_tid_valid = dummy_table_am_tuple_tid_valid,
+	.tuple_satisfies_snapshot = dummy_table_am_tuple_satisfies_snapshot,
+	.index_delete_tuples = dummy_table_am_index_delete_tuples,
+
+	.relation_set_new_filelocator = dummy_table_am_relation_set_new_filelocator,
+	.relation_nontransactional_truncate = dummy_table_am_relation_nontransactional_truncate,
+	.relation_copy_data = dummy_table_am_copy_data,
+	.relation_copy_for_cluster = dummy_table_am_copy_for_cluster,
+	.relation_vacuum = dummy_table_am_vacuum,
+	.scan_analyze_next_block = dummy_table_am_scan_analyze_next_block,
+	.scan_analyze_next_tuple = dummy_table_am_scan_analyze_next_tuple,
+	.index_build_range_scan = dummy_table_am_index_build_range_scan,
+	.index_validate_scan = dummy_table_am_index_validate_scan,
+
+	.relation_size = dummy_table_am_relation_size,
+	.relation_needs_toast_table = dummy_table_am_relation_needs_toast_table,
+
+	.relation_estimate_size = dummy_table_am_estimate_rel_size,
+
+	.scan_bitmap_next_block = dummy_table_am_scan_bitmap_next_block,
+	.scan_bitmap_next_tuple = dummy_table_am_scan_bitmap_next_tuple,
+	.scan_sample_next_block = dummy_table_am_scan_sample_next_block,
+	.scan_sample_next_tuple = dummy_table_am_scan_sample_next_tuple
+};
+
+/*
+ * Table AM handler function: returns TableAmRoutine with access method
+ * parameters and callbacks.
+ */
+Datum
+dummy_table_am_handler(PG_FUNCTION_ARGS)
+{
+	PG_RETURN_POINTER(&dummy_table_am_methods);
+}
diff --git a/src/test/modules/dummy_table_am/dummy_table_am.control b/src/test/modules/dummy_table_am/dummy_table_am.control
new file mode 100644
index 0000000000..08f2f868d4
--- /dev/null
+++ b/src/test/modules/dummy_table_am/dummy_table_am.control
@@ -0,0 +1,5 @@
+# dummy_table_am extension
+comment = 'dummy_table_am - table access method template'
+default_version = '1.0'
+module_pathname = '$libdir/dummy_table_am'
+relocatable = true
diff --git a/src/test/modules/dummy_table_am/expected/dummy_table_am.out b/src/test/modules/dummy_table_am/expected/dummy_table_am.out
new file mode 100644
index 0000000000..76029a30ae
--- /dev/null
+++ b/src/test/modules/dummy_table_am/expected/dummy_table_am.out
@@ -0,0 +1,207 @@
+-- Tests for dummy table access method
+CREATE EXTENSION dummy_table_am;
+CREATE TABLE dummy_table (a int, b text) USING dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+-- Will error out
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INFO:  dummy_table_am_estimate_rel_size
+ERROR:  dummy_table_am_index_build_range_scan
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+INSERT INTO dummy_table VALUES (1, 'dummy');
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+INSERT INTO dummy_table VALUES (1, 'dummy'), (2, 'dummy');
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+INFO:  dummy_table_am_tuple_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+UPDATE dummy_table SET a = 0 WHERE a = 1;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+-- Update without WHERE
+UPDATE dummy_table SET a = 0, b = NULL;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+DELETE FROM dummy_table WHERE a = 1;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+TRUNCATE dummy_table;
+INFO:  dummy_table_am_relation_set_new_filelocator
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+VACUUM dummy_table;
+INFO:  dummy_table_am_vacuum
+ANALYZE dummy_table;
+INFO:  dummy_table_am_relation_size
+INFO:  dummy_table_am_relation_size
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_end
+VACUUM (FULL) dummy_table;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_copy_for_cluster
+INFO:  dummy_table_am_relation_size
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+COPY dummy_table TO STDOUT;
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+COPY dummy_table (a, b) FROM STDIN;
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_multi_insert
+INFO:  dummy_table_am_finish_bulk_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b 
+---+---
+(0 rows)
+
+ALTER TABLE dummy_table ADD COLUMN c BOOLEAN DEFAULT true;
+INFO:  dummy_table_am_relation_needs_toast_table
+INSERT INTO dummy_table VALUES (1, 'dummy');
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b | c 
+---+---+---
+(0 rows)
+
+-- ALTER TABLE SET ACCESS METHOD
+ALTER TABLE dummy_table SET ACCESS METHOD heap;
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INSERT INTO dummy_table VALUES (1, 'heap', false);
+SELECT * FROM dummy_table;
+ a |  b   | c 
+---+------+---
+ 1 | heap | f
+(1 row)
+
+-- Will error out, the index must be dropped
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+INFO:  dummy_table_am_finish_bulk_insert
+INFO:  dummy_table_am_estimate_rel_size
+ERROR:  dummy_table_am_index_build_range_scan
+DROP INDEX dummy_table_idx;
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+INFO:  dummy_table_am_relation_set_new_filelocator
+INFO:  dummy_table_am_relation_needs_toast_table
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_tuple_insert
+INFO:  dummy_table_am_finish_bulk_insert
+SELECT * FROM dummy_table;
+INFO:  dummy_table_am_estimate_rel_size
+INFO:  dummy_table_am_slot_callbacks
+INFO:  dummy_table_am_scan_begin
+INFO:  dummy_table_am_scan_getnextslot
+INFO:  dummy_table_am_scan_end
+ a | b | c 
+---+---+---
+(0 rows)
+
+-- Clean up
+DROP TABLE dummy_table;
diff --git a/src/test/modules/dummy_table_am/meson.build b/src/test/modules/dummy_table_am/meson.build
new file mode 100644
index 0000000000..93f9108f29
--- /dev/null
+++ b/src/test/modules/dummy_table_am/meson.build
@@ -0,0 +1,33 @@
+# Copyright (c) 2022-2023, PostgreSQL Global Development Group
+
+dummy_table_am_sources = files(
+  'dummy_table_am.c',
+)
+
+if host_system == 'windows'
+  dummy_table_am_sources += rc_lib_gen.process(win32ver_rc, extra_args: [
+    '--NAME', 'dummy_table_am',
+    '--FILEDESC', 'dummy_table_am - table access method template',])
+endif
+
+dummy_table_am = shared_module('dummy_table_am',
+  dummy_table_am_sources,
+  kwargs: pg_test_mod_args,
+)
+test_install_libs += dummy_table_am
+
+test_install_data += files(
+  'dummy_table_am.control',
+  'dummy_table_am--1.0.sql',
+)
+
+tests += {
+  'name': 'dummy_table_am',
+  'sd': meson.current_source_dir(),
+  'bd': meson.current_build_dir(),
+  'regress': {
+    'sql': [
+      'dummy_table_am',
+    ],
+  },
+}
diff --git a/src/test/modules/dummy_table_am/sql/dummy_table_am.sql b/src/test/modules/dummy_table_am/sql/dummy_table_am.sql
new file mode 100644
index 0000000000..f17c169255
--- /dev/null
+++ b/src/test/modules/dummy_table_am/sql/dummy_table_am.sql
@@ -0,0 +1,55 @@
+-- Tests for dummy table access method
+CREATE EXTENSION dummy_table_am;
+CREATE TABLE dummy_table (a int, b text) USING dummy_table_am;
+-- Will error out
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+SELECT * FROM dummy_table;
+
+INSERT INTO dummy_table VALUES (1, 'dummy');
+SELECT * FROM dummy_table;
+
+INSERT INTO dummy_table VALUES (1, 'dummy'), (2, 'dummy');
+SELECT * FROM dummy_table;
+
+UPDATE dummy_table SET a = 0 WHERE a = 1;
+SELECT * FROM dummy_table;
+
+-- Update without WHERE
+UPDATE dummy_table SET a = 0, b = NULL;
+SELECT * FROM dummy_table;
+
+DELETE FROM dummy_table WHERE a = 1;
+SELECT * FROM dummy_table;
+
+TRUNCATE dummy_table;
+SELECT * FROM dummy_table;
+
+VACUUM dummy_table;
+ANALYZE dummy_table;
+VACUUM (FULL) dummy_table;
+SELECT * FROM dummy_table;
+
+COPY dummy_table TO STDOUT;
+COPY dummy_table (a, b) FROM STDIN;
+1	dummy
+\.
+SELECT * FROM dummy_table;
+
+ALTER TABLE dummy_table ADD COLUMN c BOOLEAN DEFAULT true;
+INSERT INTO dummy_table VALUES (1, 'dummy');
+SELECT * FROM dummy_table;
+
+-- ALTER TABLE SET ACCESS METHOD
+ALTER TABLE dummy_table SET ACCESS METHOD heap;
+CREATE INDEX dummy_table_idx ON dummy_table (a);
+INSERT INTO dummy_table VALUES (1, 'heap', false);
+SELECT * FROM dummy_table;
+
+-- Will error out, the index must be dropped
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+DROP INDEX dummy_table_idx;
+ALTER TABLE dummy_table SET ACCESS METHOD dummy_table_am;
+SELECT * FROM dummy_table;
+
+-- Clean up
+DROP TABLE dummy_table;
diff --git a/src/test/modules/meson.build b/src/test/modules/meson.build
index 17d369e378..84460f27b4 100644
--- a/src/test/modules/meson.build
+++ b/src/test/modules/meson.build
@@ -5,6 +5,7 @@ subdir('commit_ts')
 subdir('delay_execution')
 subdir('dummy_index_am')
 subdir('dummy_seclabel')
+subdir('dummy_table_am')
 subdir('ldap_password_func')
 subdir('libpq_pipeline')
 subdir('plsample')
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 26+ messages in thread

* Re: Add test module for Table Access Method
@ 2023-09-28 01:08  Michael Paquier <[email protected]>
  parent: Fabrízio de Royes Mello <[email protected]>
  1 sibling, 0 replies; 26+ messages in thread

From: Michael Paquier @ 2023-09-28 01:08 UTC (permalink / raw)
  To: Fabrízio de Royes Mello <[email protected]>; +Cc: pgsql-hackers

On Sat, Jun 03, 2023 at 07:42:36PM -0400, Fabrízio de Royes Mello wrote:
> So in order to improve things a bit in this area I'm proposing to add a
> test module for Table Access Method similar what we already have for Index
> Access Method.
> 
> This code is based on the "blackhole_am" implemented by Michael Paquier:
> https://github.com/michaelpq/pg_plugins/tree/main/blackhole_am

dummy_index_am has included from the start additional coverage for the
various internal add_*_reloption routines, that were never covered in
the core tree.  Except if I am missing something, I am not seeing some
of the extra usefulness for the patch you've sent here.
--
Michael


Attachments:

  [application/pgp-signature] signature.asc (833B, ../../[email protected]/2-signature.asc)
  download

^ permalink  raw  reply  [nested|flat] 26+ messages in thread

end of thread, other threads:[~2023-09-28 01:08 UTC | newest]

Thread overview: 26+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH v5 1/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2021-02-02 17:03 [PATCH 2/2] make LogwrtResult atomic Alvaro Herrera <[email protected]>
2023-06-03 23:42 Add test module for Table Access Method Fabrízio de Royes Mello <[email protected]>
2023-06-05 16:24 ` Re: Add test module for Table Access Method Fabrízio de Royes Mello <[email protected]>
2023-09-28 01:08 ` Re: Add test module for Table Access Method Michael Paquier <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox