agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH] hint squash commit
53+ messages / 5 participants
[nested] [flat]

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 03:25  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 03:25 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +++++-
 src/backend/access/transam/xlog.c        |  4 ++++
 src/backend/access/transam/xloginsert.c  | 26 ++++++++++++++++++++++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 18 ++++++++++++++++
 src/include/access/xloginsert.h          |  2 ++
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..eebca0fc16 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,32 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ *
+ * Callable while holding just share lock on the buffer content.
+ *
+ * The LSN of the inserted wal record is returned if we had to write,
+ * InvalidXLogRecPtr otherwise.
+ *
+ * It is possible that multiple concurrent backends could attempt to write WAL
+ * records. In that case, multiple copies of the same block would be recorded
+ * in separate WAL records by different backends, though that is still OK from
+ * a correctness perspective.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..fd59f9fc1d 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3865,6 +3865,24 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		{
 			dirtied = true;		/* Means "will be dirtied by this action" */
 
+			/*
+			 * If the buffer is clean, and lsn is valid, it must be the
+			 * first hint bit change of the checkpoint (the old page lsn
+			 * is earlier than the RedoRecPtr).  If the page is clean and
+			 * the lsn is invalid, the page must have been already written
+			 * during this checkpoint, and this is a new hint bit change.
+			 * Such page changes usually don't create new page lsns, but we
+			 * need one for cluster file encryption because the lsn is used as
+			 * part of the nonce, and we don't want to reencrypt the page
+			 * with the hint bit changes using the same nonce as the previous
+			 * writes during this checkpoint.  (It would expose the hint bit
+			 * change locations.)  Therefore we create a simple WAL record
+			 * to advance the lsn, which can can then be assigned to the
+			 * page.
+			 */
+			if (XLogRecPtrIsInvalid(lsn) /* XXX && file_encryption_method != 0 */)
+				lsn = XLogLSNForHint();
+
 			/*
 			 * Set the page LSN if we wrote a backup block. We aren't supposed
 			 * to set this when only holding a share lock but as long as we
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--ikeVEW9yuYc//A+q--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* [PATCH] hint squash commit
@ 2021-02-04 18:43  Bruce Momjian <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Bruce Momjian @ 2021-02-04 18:43 UTC (permalink / raw)

---
 src/backend/access/rmgrdesc/xlogdesc.c   |  6 +-
 src/backend/access/transam/xlog.c        |  4 ++
 src/backend/access/transam/xloginsert.c  | 16 +++++
 src/backend/replication/logical/decode.c |  1 +
 src/backend/storage/buffer/bufmgr.c      | 89 ++++++++++++++++--------
 src/include/access/xloginsert.h          |  2 +
 src/include/catalog/pg_control.h         |  1 +
 7 files changed, 90 insertions(+), 29 deletions(-)

diff --git a/src/backend/access/rmgrdesc/xlogdesc.c b/src/backend/access/rmgrdesc/xlogdesc.c
index 92cc7ea073..16a967bb71 100644
--- a/src/backend/access/rmgrdesc/xlogdesc.c
+++ b/src/backend/access/rmgrdesc/xlogdesc.c
@@ -80,7 +80,8 @@ xlog_desc(StringInfo buf, XLogReaderState *record)
 
 		appendStringInfoString(buf, xlrec->rp_name);
 	}
-	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT)
+	else if (info == XLOG_FPI || info == XLOG_FPI_FOR_HINT ||
+			 info == XLOG_HINT_LSN)
 	{
 		/* no further information to print */
 	}
@@ -185,6 +186,9 @@ xlog_identify(uint8 info)
 		case XLOG_FPI_FOR_HINT:
 			id = "FPI_FOR_HINT";
 			break;
+		case XLOG_HINT_LSN:
+			id = "HINT_LSN";
+			break;
 	}
 
 	return id;
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f03bd473e2..f67316ee07 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -10260,6 +10260,10 @@ xlog_redo(XLogReaderState *record)
 			UnlockReleaseBuffer(buffer);
 		}
 	}
+	else if (info == XLOG_HINT_LSN)
+	{
+		/* nothing to do here */
+	}
 	else if (info == XLOG_BACKUP_END)
 	{
 		XLogRecPtr	startpoint;
diff --git a/src/backend/access/transam/xloginsert.c b/src/backend/access/transam/xloginsert.c
index 7052dc245e..7635a3d44d 100644
--- a/src/backend/access/transam/xloginsert.c
+++ b/src/backend/access/transam/xloginsert.c
@@ -980,6 +980,22 @@ XLogSaveBufferForHint(Buffer buffer, bool buffer_std)
 	return recptr;
 }
 
+/*
+ * On the first hint bit change during a checkpoint, XLogSaveBufferForHint()
+ * writes a full page image to the WAL and returns a new LSN which can be
+ * assigned to the page.  Cluster file encryption needs a new LSN for
+ * every hint bit page write because the LSN is used in the nonce, and
+ * the nonce must be unique.  Therefore, this routine just advances the LSN
+ * so the page can be assigned a new LSN.
+ */
+XLogRecPtr
+XLogLSNForHint(void)
+{
+	XLogBeginInsert();
+
+	return XLogInsert(RM_XLOG_ID, XLOG_HINT_LSN);
+}
+
 /*
  * Write a WAL record containing a full image of a page. Caller is responsible
  * for writing the page to disk after calling this routine.
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index afa1df00d0..2ada89c6b1 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -223,6 +223,7 @@ DecodeXLogOp(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
 		case XLOG_FPW_CHANGE:
 		case XLOG_FPI_FOR_HINT:
 		case XLOG_FPI:
+		case XLOG_HINT_LSN:
 			break;
 		default:
 			elog(ERROR, "unexpected RM_XLOG_ID record type: %u", info);
diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c
index 561c212092..b9f3f76798 100644
--- a/src/backend/storage/buffer/bufmgr.c
+++ b/src/backend/storage/buffer/bufmgr.c
@@ -3810,13 +3810,14 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 		 * If we need to protect hint bit updates from torn writes, WAL-log a
 		 * full page image of the page. This full page image is only necessary
 		 * if the hint bit update is the first change to the page since the
-		 * last checkpoint.
+		 * last checkpoint.  If cluster file encryption is enabled, we also
+		 * need to generate new page LSNs for non-first-page writes during
+		 * a checkpoint.
 		 *
 		 * We don't check full_page_writes here because that logic is included
 		 * when we call XLogInsert() since the value changes dynamically.
 		 */
-		if (XLogHintBitIsNeeded() &&
-			(pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT))
+		if (XLogHintBitIsNeeded())
 		{
 			/*
 			 * If we must not write WAL, due to a relfilenode-specific
@@ -3826,35 +3827,67 @@ MarkBufferDirtyHint(Buffer buffer, bool buffer_std)
 			 *
 			 * See src/backend/storage/page/README for longer discussion.
 			 */
-			if (RecoveryInProgress() ||
-				RelFileNodeSkippingWAL(bufHdr->tag.rnode))
+			if (((pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+				 /* XXX || file_encryption_method != 0 */) &&
+				(RecoveryInProgress() ||
+				 RelFileNodeSkippingWAL(bufHdr->tag.rnode)))
 				return;
 
+			if (pg_atomic_read_u32(&bufHdr->state) & BM_PERMANENT)
+			{
+				/*
+				 * If the block is already dirty because we either made a change
+				 * or set a hint already, then we don't need to write a full page
+				 * image.  Note that aggressive cleaning of blocks dirtied by hint
+				 * bit setting would increase the call rate. Bulk setting of hint
+				 * bits would reduce the call rate...
+				 *
+				 * We must issue the WAL record before we mark the buffer dirty.
+				 * Otherwise we might write the page before we write the WAL. That
+				 * causes a race condition, since a checkpoint might occur between
+				 * writing the WAL record and marking the buffer dirty. We solve
+				 * that with a kluge, but one that is already in use during
+				 * transaction commit to prevent race conditions. Basically, we
+				 * simply prevent the checkpoint WAL record from being written
+				 * until we have marked the buffer dirty. We don't start the
+				 * checkpoint flush until we have marked dirty, so our checkpoint
+				 * must flush the change to disk successfully or the checkpoint
+				 * never gets written, so crash recovery will fix.
+				 *
+				 * It's possible we may enter here without an xid, so it is
+				 * essential that CreateCheckpoint waits for virtual transactions
+				 * rather than full transactionids.
+				 */
+				MyProc->delayChkpt = delayChkpt = true;
+				lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			}
+
 			/*
-			 * If the block is already dirty because we either made a change
-			 * or set a hint already, then we don't need to write a full page
-			 * image.  Note that aggressive cleaning of blocks dirtied by hint
-			 * bit setting would increase the call rate. Bulk setting of hint
-			 * bits would reduce the call rate...
-			 *
-			 * We must issue the WAL record before we mark the buffer dirty.
-			 * Otherwise we might write the page before we write the WAL. That
-			 * causes a race condition, since a checkpoint might occur between
-			 * writing the WAL record and marking the buffer dirty. We solve
-			 * that with a kluge, but one that is already in use during
-			 * transaction commit to prevent race conditions. Basically, we
-			 * simply prevent the checkpoint WAL record from being written
-			 * until we have marked the buffer dirty. We don't start the
-			 * checkpoint flush until we have marked dirty, so our checkpoint
-			 * must flush the change to disk successfully or the checkpoint
-			 * never gets written, so crash recovery will fix.
-			 *
-			 * It's possible we may enter here without an xid, so it is
-			 * essential that CreateCheckpoint waits for virtual transactions
-			 * rather than full transactionids.
+			 * For cluster file encryption, we generate a new page LSN
+			 * for hint-bit non-permanent and non-first page writes.
 			 */
-			MyProc->delayChkpt = delayChkpt = true;
-			lsn = XLogSaveBufferForHint(buffer, buffer_std);
+			if (/* XXX file_encryption_method != 0 && */
+				XLogRecPtrIsInvalid(lsn) &&
+				/* XXX can we check BM_DIRTY without a page lock? */
+				!(pg_atomic_read_u32(&bufHdr->state) & BM_DIRTY))
+			{
+				/*
+				 * If the buffer is clean, and lsn is valid, it must be the
+				 * first hint bit change of the checkpoint (the old page lsn
+				 * is earlier than the RedoRecPtr).  If the page is clean and
+				 * the lsn is invalid, the page must have been already written
+				 * during this checkpoint, and this is a new hint bit change.
+				 * Such page changes usually don't create new page lsns, but we
+				 * need one for cluster file encryption because the lsn is used as
+				 * part of the nonce, and we don't want to reencrypt the page
+				 * with the hint bit changes using the same nonce as the previous
+				 * writes during this checkpoint.  (It would expose the hint bit
+				 * change locations.)  Therefore we create a simple WAL record
+				 * to advance the lsn, which can can then be assigned to the
+				 * page below.
+				 */
+				lsn = XLogLSNForHint();
+			}
 		}
 
 		buf_state = LockBufHdr(bufHdr);
diff --git a/src/include/access/xloginsert.h b/src/include/access/xloginsert.h
index f1d8c39edf..a066f65229 100644
--- a/src/include/access/xloginsert.h
+++ b/src/include/access/xloginsert.h
@@ -61,6 +61,8 @@ extern void log_newpage_range(Relation rel, ForkNumber forkNum,
 							  BlockNumber startblk, BlockNumber endblk, bool page_std);
 extern XLogRecPtr XLogSaveBufferForHint(Buffer buffer, bool buffer_std);
 
+extern XLogRecPtr XLogLSNForHint(void);
+
 extern void InitXLogInsert(void);
 
 #endif							/* XLOGINSERT_H */
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index e3f48158ce..36ac7dfbb8 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -76,6 +76,7 @@ typedef struct CheckPoint
 #define XLOG_END_OF_RECOVERY			0x90
 #define XLOG_FPI_FOR_HINT				0xA0
 #define XLOG_FPI						0xB0
+#define XLOG_HINT_LSN					0xC0
 
 
 /*
-- 
2.20.1


--T4sUOijqQbZv57TR--





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-24 16:03  Jelte Fennema <[email protected]>
  0 siblings, 2 replies; 53+ messages in thread

From: Jelte Fennema @ 2023-10-24 16:03 UTC (permalink / raw)
  To: pgsql-hackers

Many usages of the foreach macro in the Postgres codebase only use the
ListCell variable to then get its value. This adds macros that
simplify iteration code for that very common use case. Instead of
passing a ListCell you can pass a variable of the type of its
contents. This IMHO improves readability of the code by reducing the
total amount of code while also essentially forcing the use of useful
variable names.

While this might seem like a small quality of life improvement, in
practice it turns out to be very nice to use. At Microsoft we have
been using macros very similar to these ones in the Citus codebase for
a long time now and we pretty much never use plain foreach anymore for
new code.

Finally, I guess there needs to be some bikeshedding on the naming. In
the Citus codebase we call them foreach_xyz instead of the
for_each_xyz naming pattern that is used in this patchset. I'm not
sure what the current stance is on if foreach should be written with
or without an underscore between for and each. Currently pg_list.h
uses both.

P.S. Similar macros for forboth/forthree are also possible, but
require an exponential macro count handle all different possibilities,
which might not be worth the effort since forboth/forthree are used
much less often than foreach. In Citus we do have 3 forboth macros
that don't require ListCell for the most common cases (foreach_ptr,
foreach_ptr_oid, foreach_int_oid). But I did not want to clutter this
patchset with that discussion.


Attachments:

  [application/octet-stream] v1-0001-Add-macros-for-looping-through-a-list-without-nee.patch (4.6K, ../../CAGECzQSwXKnxGwW1_Q5JE+8Ja20kyAbhBHO04vVrQsLcDciwXA@mail.gmail.com/2-v1-0001-Add-macros-for-looping-through-a-list-without-nee.patch)
  download | inline diff:
From 0effd91ab9ee3b3c678f1d128e487b9a728c8588 Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 16:46:08 +0200
Subject: [PATCH v1 1/2] Add macros for looping through a list without needing
 a ListCell

Many usages of the foreach macro only use the ListCell variable to get
its contents. This adds macros that simplify iteration code for that
common use case.
---
 src/include/nodes/pg_list.h | 92 ++++++++++++++++++++++++++++++++++---
 1 file changed, 86 insertions(+), 6 deletions(-)

diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index 529a382d284..af1ecc5e394 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -383,13 +383,15 @@ lnext(const List *l, const ListCell *c)
  *	  delete the current list element from the List associated with a
  *	  surrounding foreach() loop, returning the new List pointer.
  *
- * This is equivalent to list_delete_cell(), but it also adjusts the foreach
- * loop's state so that no list elements will be missed.  Do not delete
- * elements from an active foreach loop's list in any other way!
+ * This is similar to list_delete_cell(), but it also works when using
+ * for_each_xyz macros that don't require passing in a "ListCell *".
+ * Furthermore it adjusts the foreach loop's state so that no list elements
+ * will be missed. Do not delete elements from an active foreach loop's list in
+ * any other way!
  */
-#define foreach_delete_current(lst, cell)	\
-	(cell##__state.i--, \
-	 (List *) (cell##__state.l = list_delete_cell(lst, cell)))
+#define foreach_delete_current(lst, var)	\
+	(var##__state.i--, \
+	 (List *) (var##__state.l = list_delete_cell(lst, &var##__state.l->elements[var##__state.i])))
 
 /*
  * foreach_current_index -
@@ -452,6 +454,84 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 	return r;
 }
 
+/*
+ * for_each_ptr -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *", just a declared pointer variable to store the
+ *	  current pointer int;
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_ptr(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = NULL, false); \
+		 var##__state.i++)
+
+/*
+ * for_each_int -
+ *	  a convenience macro which loops through a list of ints without needing a
+ *	  "ListCell *", just a declared int variable to store the current int in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_int(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_int(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * foreach_oid -
+ *	  a convenience macro which loops through an oid list without needing a
+ *	  "ListCell *", just a declared Oid variable to store the current oid in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_oid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_oid(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * foreach_xid -
+ *	  a convenience macro which loops through an xid list without needing a
+ *	  "ListCell *", just a declared TransactionId variable to store the current
+ *	  xid in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_xid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_xid(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * for_each_node -
+ *	  a convenience macro which loops through a node list without needing a
+ *	  "ListCell *", just a declared pointer variable to store the pointer of
+ *	  the current node in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_node(type, var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_node(type, &var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
 /*
  * forboth -
  *	  a convenience macro for advancing through two linked lists

base-commit: 00d7fb5e2e39198387ae00af8dd18b787b6a4d63
-- 
2.34.1



  [application/octet-stream] v1-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch (7.3K, ../../CAGECzQSwXKnxGwW1_Q5JE+8Ja20kyAbhBHO04vVrQsLcDciwXA@mail.gmail.com/3-v1-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch)
  download | inline diff:
From 12bce0f7d0e999f048b1847d3d8513ea2daaa1ed Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 17:13:20 +0200
Subject: [PATCH v1 2/2] Use new for_each_xyz macros in a few places

This starts using each of the newly added for_each_xyz macros in at
least one place. This shows how they improve readability by reducing the
amount of boilerplate code.
---
 src/backend/executor/execExpr.c             | 11 ++++----
 src/backend/executor/nodeSubplan.c          | 28 +++++++++------------
 src/backend/replication/logical/relation.c  |  5 ++--
 src/backend/replication/logical/tablesync.c |  6 ++---
 src/backend/replication/pgoutput/pgoutput.c |  8 +++---
 5 files changed, 25 insertions(+), 33 deletions(-)

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index 2c62b0c9c84..e73261ec445 100644
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -216,7 +216,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	ExprState  *state;
 	ExprEvalStep scratch = {0};
 	List	   *adjust_jumps = NIL;
-	ListCell   *lc;
+	Expr	   *node;
+	int			jump;
 
 	/* short-circuit (here and in ExecQual) for empty restriction list */
 	if (qual == NIL)
@@ -250,10 +251,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	scratch.resvalue = &state->resvalue;
 	scratch.resnull = &state->resnull;
 
-	foreach(lc, qual)
+	for_each_ptr(node, qual)
 	{
-		Expr	   *node = (Expr *) lfirst(lc);
-
 		/* first evaluate expression */
 		ExecInitExprRec(node, state, &state->resvalue, &state->resnull);
 
@@ -265,9 +264,9 @@ ExecInitQual(List *qual, PlanState *parent)
 	}
 
 	/* adjust jump targets */
-	foreach(lc, adjust_jumps)
+	for_each_int(jump, adjust_jumps)
 	{
-		ExprEvalStep *as = &state->steps[lfirst_int(lc)];
+		ExprEvalStep *as = &state->steps[jump];
 
 		Assert(as->opcode == EEOP_QUAL);
 		Assert(as->d.qualexpr.jumpdone == -1);
diff --git a/src/backend/executor/nodeSubplan.c b/src/backend/executor/nodeSubplan.c
index c136f75ac24..8f88f289269 100644
--- a/src/backend/executor/nodeSubplan.c
+++ b/src/backend/executor/nodeSubplan.c
@@ -307,7 +307,7 @@ ExecScanSubPlan(SubPlanState *node,
 		Datum		rowresult;
 		bool		rownull;
 		int			col;
-		ListCell   *plst;
+		int			paramid;
 
 		if (subLinkType == EXISTS_SUBLINK)
 		{
@@ -367,9 +367,8 @@ ExecScanSubPlan(SubPlanState *node,
 			 * Now set all the setParam params from the columns of the tuple
 			 */
 			col = 1;
-			foreach(plst, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(plst);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -412,9 +411,8 @@ ExecScanSubPlan(SubPlanState *node,
 		 * combining expression.
 		 */
 		col = 1;
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -480,10 +478,11 @@ ExecScanSubPlan(SubPlanState *node,
 		}
 		else if (subLinkType == MULTIEXPR_SUBLINK)
 		{
+			int			paramid;
+
 			/* We don't care about function result, but set the setParams */
-			foreach(l, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(l);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -604,16 +603,15 @@ buildSubPlanHash(SubPlanState *node, ExprContext *econtext)
 		 slot = ExecProcNode(planstate))
 	{
 		int			col = 1;
-		ListCell   *plst;
 		bool		isnew;
+		int			paramid;
 
 		/*
 		 * Load up the Params representing the raw sub-select outputs, then
 		 * form the projection tuple to store in the hashtable.
 		 */
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
@@ -880,11 +878,10 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 	if (subplan->setParam != NIL && subplan->parParam == NIL &&
 		subplan->subLinkType != CTE_SUBLINK)
 	{
-		ListCell   *lst;
+		int			paramid;
 
-		foreach(lst, subplan->setParam)
+		for_each_int(paramid, subplan->setParam)
 		{
-			int			paramid = lfirst_int(lst);
 			ParamExecData *prm = &(estate->es_param_exec_vals[paramid]);
 
 			prm->execPlan = sstate;
@@ -906,7 +903,7 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		List	   *oplist,
 				   *lefttlist,
 				   *righttlist;
-		ListCell   *l;
+		OpExpr	   *opexpr;
 
 		/* We need a memory context to hold the hash table(s) */
 		sstate->hashtablecxt =
@@ -966,9 +963,8 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		cross_eq_funcoids = (Oid *) palloc(ncols * sizeof(Oid));
 
 		i = 1;
-		foreach(l, oplist)
+		for_each_node(OpExpr, opexpr, oplist)
 		{
-			OpExpr	   *opexpr = lfirst_node(OpExpr, l);
 			Expr	   *expr;
 			TargetEntry *tle;
 			Oid			rhs_eq_oper;
diff --git a/src/backend/replication/logical/relation.c b/src/backend/replication/logical/relation.c
index d62eefed138..f8faceb57ce 100644
--- a/src/backend/replication/logical/relation.c
+++ b/src/backend/replication/logical/relation.c
@@ -747,11 +747,10 @@ static Oid
 FindUsableIndexForReplicaIdentityFull(Relation localrel, AttrMap *attrmap)
 {
 	List	   *idxlist = RelationGetIndexList(localrel);
-	ListCell   *lc;
+	Oid			idxoid;
 
-	foreach(lc, idxlist)
+	for_each_oid(idxoid, idxlist)
 	{
-		Oid			idxoid = lfirst_oid(lc);
 		bool		isUsableIdx;
 		Relation	idxRel;
 		IndexInfo  *idxInfo;
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 37a0abe2f4d..c82d05a642e 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	ListCell   *lc;
+	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,10 +453,8 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	foreach(lc, table_states_not_ready)
+	for_each_ptr(rstate, table_states_not_ready)
 	{
-		SubscriptionRelState *rstate = (SubscriptionRelState *) lfirst(lc);
-
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
 			/*
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index c1c66848f36..ec6df59bedc 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -2229,7 +2229,7 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 {
 	HASH_SEQ_STATUS hash_seq;
 	RelationSyncEntry *entry;
-	ListCell   *lc;
+	TransactionId streamed_txn;
 
 	Assert(RelationSyncCache != NULL);
 
@@ -2242,15 +2242,15 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 		 * corresponding schema and we don't need to send it unless there is
 		 * any invalidation for that relation.
 		 */
-		foreach(lc, entry->streamed_txns)
+		for_each_xid(streamed_txn, entry->streamed_txns)
 		{
-			if (xid == lfirst_xid(lc))
+			if (xid == streamed_txn)
 			{
 				if (is_commit)
 					entry->schema_sent = true;
 
 				entry->streamed_txns =
-					foreach_delete_current(entry->streamed_txns, lc);
+					foreach_delete_current(entry->streamed_txns, streamed_txn);
 				break;
 			}
 		}
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-24 16:47  Nathan Bossart <[email protected]>
  parent: Jelte Fennema <[email protected]>
  1 sibling, 1 reply; 53+ messages in thread

From: Nathan Bossart @ 2023-10-24 16:47 UTC (permalink / raw)
  To: Jelte Fennema <[email protected]>; +Cc: pgsql-hackers

On Tue, Oct 24, 2023 at 06:03:48PM +0200, Jelte Fennema wrote:
> Many usages of the foreach macro in the Postgres codebase only use the
> ListCell variable to then get its value. This adds macros that
> simplify iteration code for that very common use case. Instead of
> passing a ListCell you can pass a variable of the type of its
> contents. This IMHO improves readability of the code by reducing the
> total amount of code while also essentially forcing the use of useful
> variable names.
> 
> While this might seem like a small quality of life improvement, in
> practice it turns out to be very nice to use. At Microsoft we have
> been using macros very similar to these ones in the Citus codebase for
> a long time now and we pretty much never use plain foreach anymore for
> new code.

This seems reasonable to me.

> Finally, I guess there needs to be some bikeshedding on the naming. In
> the Citus codebase we call them foreach_xyz instead of the
> for_each_xyz naming pattern that is used in this patchset. I'm not
> sure what the current stance is on if foreach should be written with
> or without an underscore between for and each. Currently pg_list.h
> uses both.

I don't have a strong opinion on the matter, but if I had to choose, I
guess I'd pick foreach_*() because these macros are most closely related to
foreach().

BTW after applying your patches, initdb began failing with the following
for me:

	TRAP: failed Assert("n >= 0 && n < list->length"), File: "list.c", Line: 770, PID: 902807

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-24 16:58  Jelte Fennema <[email protected]>
  parent: Nathan Bossart <[email protected]>
  0 siblings, 2 replies; 53+ messages in thread

From: Jelte Fennema @ 2023-10-24 16:58 UTC (permalink / raw)
  To: Nathan Bossart <[email protected]>; +Cc: pgsql-hackers

On Tue, 24 Oct 2023 at 18:47, Nathan Bossart <[email protected]> wrote:
> BTW after applying your patches, initdb began failing with the following
> for me:
>
>         TRAP: failed Assert("n >= 0 && n < list->length"), File: "list.c", Line: 770, PID: 902807

Oh oops... That was an off by one error in the modified
foreach_delete_current implementation.
Attached is a fixed version.


Attachments:

  [application/octet-stream] v2-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch (7.3K, ../../CAGECzQQG7VX3LU4ccwygSeoVSM+UdSr+O8Mt40f0WEuoWV+ohg@mail.gmail.com/2-v2-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch)
  download | inline diff:
From b295b8779a4c65d3c9f9b489264d4cbac6ac50db Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 17:13:20 +0200
Subject: [PATCH v2 2/2] Use new for_each_xyz macros in a few places

This starts using each of the newly added for_each_xyz macros in at
least one place. This shows how they improve readability by reducing the
amount of boilerplate code.
---
 src/backend/executor/execExpr.c             | 11 ++++----
 src/backend/executor/nodeSubplan.c          | 28 +++++++++------------
 src/backend/replication/logical/relation.c  |  5 ++--
 src/backend/replication/logical/tablesync.c |  6 ++---
 src/backend/replication/pgoutput/pgoutput.c |  8 +++---
 5 files changed, 25 insertions(+), 33 deletions(-)

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index 2c62b0c9c84..e73261ec445 100644
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -216,7 +216,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	ExprState  *state;
 	ExprEvalStep scratch = {0};
 	List	   *adjust_jumps = NIL;
-	ListCell   *lc;
+	Expr	   *node;
+	int			jump;
 
 	/* short-circuit (here and in ExecQual) for empty restriction list */
 	if (qual == NIL)
@@ -250,10 +251,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	scratch.resvalue = &state->resvalue;
 	scratch.resnull = &state->resnull;
 
-	foreach(lc, qual)
+	for_each_ptr(node, qual)
 	{
-		Expr	   *node = (Expr *) lfirst(lc);
-
 		/* first evaluate expression */
 		ExecInitExprRec(node, state, &state->resvalue, &state->resnull);
 
@@ -265,9 +264,9 @@ ExecInitQual(List *qual, PlanState *parent)
 	}
 
 	/* adjust jump targets */
-	foreach(lc, adjust_jumps)
+	for_each_int(jump, adjust_jumps)
 	{
-		ExprEvalStep *as = &state->steps[lfirst_int(lc)];
+		ExprEvalStep *as = &state->steps[jump];
 
 		Assert(as->opcode == EEOP_QUAL);
 		Assert(as->d.qualexpr.jumpdone == -1);
diff --git a/src/backend/executor/nodeSubplan.c b/src/backend/executor/nodeSubplan.c
index c136f75ac24..8f88f289269 100644
--- a/src/backend/executor/nodeSubplan.c
+++ b/src/backend/executor/nodeSubplan.c
@@ -307,7 +307,7 @@ ExecScanSubPlan(SubPlanState *node,
 		Datum		rowresult;
 		bool		rownull;
 		int			col;
-		ListCell   *plst;
+		int			paramid;
 
 		if (subLinkType == EXISTS_SUBLINK)
 		{
@@ -367,9 +367,8 @@ ExecScanSubPlan(SubPlanState *node,
 			 * Now set all the setParam params from the columns of the tuple
 			 */
 			col = 1;
-			foreach(plst, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(plst);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -412,9 +411,8 @@ ExecScanSubPlan(SubPlanState *node,
 		 * combining expression.
 		 */
 		col = 1;
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -480,10 +478,11 @@ ExecScanSubPlan(SubPlanState *node,
 		}
 		else if (subLinkType == MULTIEXPR_SUBLINK)
 		{
+			int			paramid;
+
 			/* We don't care about function result, but set the setParams */
-			foreach(l, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(l);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -604,16 +603,15 @@ buildSubPlanHash(SubPlanState *node, ExprContext *econtext)
 		 slot = ExecProcNode(planstate))
 	{
 		int			col = 1;
-		ListCell   *plst;
 		bool		isnew;
+		int			paramid;
 
 		/*
 		 * Load up the Params representing the raw sub-select outputs, then
 		 * form the projection tuple to store in the hashtable.
 		 */
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
@@ -880,11 +878,10 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 	if (subplan->setParam != NIL && subplan->parParam == NIL &&
 		subplan->subLinkType != CTE_SUBLINK)
 	{
-		ListCell   *lst;
+		int			paramid;
 
-		foreach(lst, subplan->setParam)
+		for_each_int(paramid, subplan->setParam)
 		{
-			int			paramid = lfirst_int(lst);
 			ParamExecData *prm = &(estate->es_param_exec_vals[paramid]);
 
 			prm->execPlan = sstate;
@@ -906,7 +903,7 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		List	   *oplist,
 				   *lefttlist,
 				   *righttlist;
-		ListCell   *l;
+		OpExpr	   *opexpr;
 
 		/* We need a memory context to hold the hash table(s) */
 		sstate->hashtablecxt =
@@ -966,9 +963,8 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		cross_eq_funcoids = (Oid *) palloc(ncols * sizeof(Oid));
 
 		i = 1;
-		foreach(l, oplist)
+		for_each_node(OpExpr, opexpr, oplist)
 		{
-			OpExpr	   *opexpr = lfirst_node(OpExpr, l);
 			Expr	   *expr;
 			TargetEntry *tle;
 			Oid			rhs_eq_oper;
diff --git a/src/backend/replication/logical/relation.c b/src/backend/replication/logical/relation.c
index d62eefed138..f8faceb57ce 100644
--- a/src/backend/replication/logical/relation.c
+++ b/src/backend/replication/logical/relation.c
@@ -747,11 +747,10 @@ static Oid
 FindUsableIndexForReplicaIdentityFull(Relation localrel, AttrMap *attrmap)
 {
 	List	   *idxlist = RelationGetIndexList(localrel);
-	ListCell   *lc;
+	Oid			idxoid;
 
-	foreach(lc, idxlist)
+	for_each_oid(idxoid, idxlist)
 	{
-		Oid			idxoid = lfirst_oid(lc);
 		bool		isUsableIdx;
 		Relation	idxRel;
 		IndexInfo  *idxInfo;
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 37a0abe2f4d..c82d05a642e 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	ListCell   *lc;
+	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,10 +453,8 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	foreach(lc, table_states_not_ready)
+	for_each_ptr(rstate, table_states_not_ready)
 	{
-		SubscriptionRelState *rstate = (SubscriptionRelState *) lfirst(lc);
-
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
 			/*
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index c1c66848f36..ec6df59bedc 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -2229,7 +2229,7 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 {
 	HASH_SEQ_STATUS hash_seq;
 	RelationSyncEntry *entry;
-	ListCell   *lc;
+	TransactionId streamed_txn;
 
 	Assert(RelationSyncCache != NULL);
 
@@ -2242,15 +2242,15 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 		 * corresponding schema and we don't need to send it unless there is
 		 * any invalidation for that relation.
 		 */
-		foreach(lc, entry->streamed_txns)
+		for_each_xid(streamed_txn, entry->streamed_txns)
 		{
-			if (xid == lfirst_xid(lc))
+			if (xid == streamed_txn)
 			{
 				if (is_commit)
 					entry->schema_sent = true;
 
 				entry->streamed_txns =
-					foreach_delete_current(entry->streamed_txns, lc);
+					foreach_delete_current(entry->streamed_txns, streamed_txn);
 				break;
 			}
 		}
-- 
2.34.1



  [application/octet-stream] v2-0001-Add-macros-for-looping-through-a-list-without-nee.patch (4.6K, ../../CAGECzQQG7VX3LU4ccwygSeoVSM+UdSr+O8Mt40f0WEuoWV+ohg@mail.gmail.com/3-v2-0001-Add-macros-for-looping-through-a-list-without-nee.patch)
  download | inline diff:
From 42e572b33bc0a7ea6b97f339f4a58b6c3c3aa729 Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 16:46:08 +0200
Subject: [PATCH v2 1/2] Add macros for looping through a list without needing
 a ListCell

Many usages of the foreach macro only use the ListCell variable to get
its contents. This adds macros that simplify iteration code for that
common use case.
---
 src/include/nodes/pg_list.h | 92 ++++++++++++++++++++++++++++++++++---
 1 file changed, 86 insertions(+), 6 deletions(-)

diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index 529a382d284..a2d98253f2a 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -383,13 +383,15 @@ lnext(const List *l, const ListCell *c)
  *	  delete the current list element from the List associated with a
  *	  surrounding foreach() loop, returning the new List pointer.
  *
- * This is equivalent to list_delete_cell(), but it also adjusts the foreach
- * loop's state so that no list elements will be missed.  Do not delete
- * elements from an active foreach loop's list in any other way!
+ * This is similar to list_delete_cell(), but it also works when using
+ * for_each_xyz macros that don't require passing in a "ListCell *".
+ * Furthermore it adjusts the foreach loop's state so that no list elements
+ * will be missed. Do not delete elements from an active foreach loop's list in
+ * any other way!
  */
-#define foreach_delete_current(lst, cell)	\
-	(cell##__state.i--, \
-	 (List *) (cell##__state.l = list_delete_cell(lst, cell)))
+#define foreach_delete_current(lst, var)	\
+	(var##__state.i--, \
+	 (List *) (var##__state.l = list_delete_cell(lst, &var##__state.l->elements[var##__state.i+1])))
 
 /*
  * foreach_current_index -
@@ -452,6 +454,84 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 	return r;
 }
 
+/*
+ * for_each_ptr -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *", just a declared pointer variable to store the
+ *	  current pointer int;
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_ptr(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = NULL, false); \
+		 var##__state.i++)
+
+/*
+ * for_each_int -
+ *	  a convenience macro which loops through a list of ints without needing a
+ *	  "ListCell *", just a declared int variable to store the current int in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_int(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_int(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * foreach_oid -
+ *	  a convenience macro which loops through an oid list without needing a
+ *	  "ListCell *", just a declared Oid variable to store the current oid in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_oid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_oid(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * foreach_xid -
+ *	  a convenience macro which loops through an xid list without needing a
+ *	  "ListCell *", just a declared TransactionId variable to store the current
+ *	  xid in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_xid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_xid(&var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
+/*
+ * for_each_node -
+ *	  a convenience macro which loops through a node list without needing a
+ *	  "ListCell *", just a declared pointer variable to store the pointer of
+ *	  the current node in.
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_node(type, var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length) ? \
+		 (var = lfirst_node(type, &var##__state.l->elements[var##__state.i]), true) : \
+		 (var = 0, false); \
+		 var##__state.i++)
+
 /*
  * forboth -
  *	  a convenience macro for advancing through two linked lists

base-commit: 00d7fb5e2e39198387ae00af8dd18b787b6a4d63
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-24 21:20  Nathan Bossart <[email protected]>
  parent: Jelte Fennema <[email protected]>
  1 sibling, 0 replies; 53+ messages in thread

From: Nathan Bossart @ 2023-10-24 21:20 UTC (permalink / raw)
  To: Jelte Fennema <[email protected]>; +Cc: pgsql-hackers

On Tue, Oct 24, 2023 at 06:58:04PM +0200, Jelte Fennema wrote:
> On Tue, 24 Oct 2023 at 18:47, Nathan Bossart <[email protected]> wrote:
>> BTW after applying your patches, initdb began failing with the following
>> for me:
>>
>>         TRAP: failed Assert("n >= 0 && n < list->length"), File: "list.c", Line: 770, PID: 902807
> 
> Oh oops... That was an off by one error in the modified
> foreach_delete_current implementation.
> Attached is a fixed version.

Thanks, that fixed it for me, too.

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-25 02:55  David Rowley <[email protected]>
  parent: Jelte Fennema <[email protected]>
  1 sibling, 1 reply; 53+ messages in thread

From: David Rowley @ 2023-10-25 02:55 UTC (permalink / raw)
  To: Jelte Fennema <[email protected]>; +Cc: Nathan Bossart <[email protected]>; pgsql-hackers

On Wed, 25 Oct 2023 at 06:00, Jelte Fennema <[email protected]> wrote:
> Attached is a fixed version.

With foreach(), we commonly do "if (lc == NULL)" at the end of loops
as a way of checking if we did "break" to terminate the loop early.
Doing the equivalent with the new macros won't be safe as the list
element's value we broke on may be set to NULL. I think it might be a
good idea to document the fact that this wouldn't be safe with the new
macros, or better yet, document the correct way to determine if we
broke out the loop early.  I imagine someone will want to do some
conversion work at some future date and it would be good if we could
avoid introducing bugs during that process.

I wonder if we should even bother setting the variable to NULL at the
end of the loop.  It feels like someone might just end up mistakenly
checking for NULLs even if we document that it's not safe.  If we left
the variable pointing to the last list element then the user of the
macro is more likely to notice their broken code. It'd also save a bit
of instruction space.

David





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-25 08:51  Alvaro Herrera <[email protected]>
  parent: Jelte Fennema <[email protected]>
  1 sibling, 1 reply; 53+ messages in thread

From: Alvaro Herrera @ 2023-10-25 08:51 UTC (permalink / raw)
  To: Jelte Fennema <[email protected]>; +Cc: pgsql-hackers

On 2023-Oct-24, Jelte Fennema wrote:

> Many usages of the foreach macro in the Postgres codebase only use the
> ListCell variable to then get its value. This adds macros that
> simplify iteration code for that very common use case. Instead of
> passing a ListCell you can pass a variable of the type of its
> contents. This IMHO improves readability of the code by reducing the
> total amount of code while also essentially forcing the use of useful
> variable names.

+1 for getting rid of useless "lc" variables.

Looking at for_each_ptr() I think it may be cleaner to follow
palloc_object()'s precedent and make it foreach_object() instead (I have
no love for the extra underscore, but I won't object to it either).  And
like foreach_node, have it receive a type name to add a cast to.

I'd imagine something like

  SubscriptionRelState *rstate;

  foreach_object(SubscriptionRelState *, rstate, table_states_not_ready)
  {

-- 
Álvaro Herrera               48°01'N 7°57'E  —  https://www.EnterpriseDB.com/





^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-25 10:05  Jelte Fennema <[email protected]>
  parent: David Rowley <[email protected]>
  0 siblings, 1 reply; 53+ messages in thread

From: Jelte Fennema @ 2023-10-25 10:05 UTC (permalink / raw)
  To: David Rowley <[email protected]>; +Cc: Nathan Bossart <[email protected]>; pgsql-hackers

On Wed, 25 Oct 2023 at 04:55, David Rowley <[email protected]> wrote:
> With foreach(), we commonly do "if (lc == NULL)" at the end of loops
> as a way of checking if we did "break" to terminate the loop early.

Afaict it's done pretty infrequently. The following crude attempt at
an estimate estimates it's only done about ~1.5% of the time a foreach
is used:
$ rg 'lc == NULL' | wc -l
13
$ rg '\bforeach\(lc,' -S | wc -l
899

> Doing the equivalent with the new macros won't be safe as the list
> element's value we broke on may be set to NULL. I think it might be a
> good idea to document the fact that this wouldn't be safe with the new
> macros, or better yet, document the correct way to determine if we
> broke out the loop early.  I imagine someone will want to do some
> conversion work at some future date and it would be good if we could
> avoid introducing bugs during that process.
>
> I wonder if we should even bother setting the variable to NULL at the
> end of the loop.  It feels like someone might just end up mistakenly
> checking for NULLs even if we document that it's not safe.  If we left
> the variable pointing to the last list element then the user of the
> macro is more likely to notice their broken code. It'd also save a bit
> of instruction space.

Makes sense. Addressed this now by mentioning this limitation and
possible workarounds in the comments of the new macros and by not
setting the loop variable to NULL/0. I don't think there's an easy way
to add this feature to these new macros natively, it's a limitation of
not having a second variable. This seems fine to me, since these new
macros are meant as an addition to foreach() instead of a complete
replacement.


Attachments:

  [application/octet-stream] v3-0001-Add-macros-for-looping-through-a-list-without-nee.patch (6.1K, ../../CAGECzQTCfAkL0L+Tir2fidx6K1Xkqd8S+pZiGvcgUP2mYL_ZZQ@mail.gmail.com/2-v3-0001-Add-macros-for-looping-through-a-list-without-nee.patch)
  download | inline diff:
From 9521630e3e738b66358e47d8334bdd053878551c Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 16:46:08 +0200
Subject: [PATCH v3 1/2] Add macros for looping through a list without needing
 a ListCell

Many usages of the foreach macro only use the ListCell variable to get
its contents. This adds macros that simplify iteration code for that
common use case.
---
 src/include/nodes/pg_list.h | 112 ++++++++++++++++++++++++++++++++++--
 1 file changed, 106 insertions(+), 6 deletions(-)

diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index 529a382d284..5e8a0318eee 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -383,13 +383,15 @@ lnext(const List *l, const ListCell *c)
  *	  delete the current list element from the List associated with a
  *	  surrounding foreach() loop, returning the new List pointer.
  *
- * This is equivalent to list_delete_cell(), but it also adjusts the foreach
- * loop's state so that no list elements will be missed.  Do not delete
- * elements from an active foreach loop's list in any other way!
+ * This is similar to list_delete_cell(), but it also works when using
+ * for_each_xyz macros that don't require passing in a "ListCell *".
+ * Furthermore it adjusts the foreach loop's state so that no list elements
+ * will be missed. Do not delete elements from an active foreach loop's list in
+ * any other way!
  */
-#define foreach_delete_current(lst, cell)	\
-	(cell##__state.i--, \
-	 (List *) (cell##__state.l = list_delete_cell(lst, cell)))
+#define foreach_delete_current(lst, var)	\
+	(var##__state.i--, \
+	 (List *) (var##__state.l = list_delete_cell(lst, &var##__state.l->elements[var##__state.i+1])))
 
 /*
  * foreach_current_index -
@@ -452,6 +454,104 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 	return r;
 }
 
+/*
+ * for_each_ptr -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *", just a declared pointer variable to store the
+ *	  current pointer int;
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_ptr(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_int -
+ *	  a convenience macro which loops through a list of ints without needing a
+ *	  "ListCell *", just a declared int variable to store the current int in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_int(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_int(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_oid -
+ *	  a convenience macro which loops through an oid list without needing a
+ *	  "ListCell *", just a declared Oid variable to store the current oid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_oid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_oid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_xid -
+ *	  a convenience macro which loops through an xid list without needing a
+ *	  "ListCell *", just a declared TransactionId variable to store the current
+ *	  xid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_xid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_xid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_node -
+ *	  a convenience macro which loops through a node list without needing a
+ *	  "ListCell *", just a declared pointer variable to store the pointer of
+ *	  the current node in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_node(type, var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_node(type, &var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
 /*
  * forboth -
  *	  a convenience macro for advancing through two linked lists

base-commit: 8f0fd47fa33720dd09ad0ae74a8a583b9780e328
-- 
2.34.1



  [application/octet-stream] v3-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch (7.3K, ../../CAGECzQTCfAkL0L+Tir2fidx6K1Xkqd8S+pZiGvcgUP2mYL_ZZQ@mail.gmail.com/3-v3-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch)
  download | inline diff:
From 89c010b0c7dd1e0879af0c6a21d91e0b75c0d38d Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 17:13:20 +0200
Subject: [PATCH v3 2/2] Use new for_each_xyz macros in a few places

This starts using each of the newly added for_each_xyz macros in at
least one place. This shows how they improve readability by reducing the
amount of boilerplate code.
---
 src/backend/executor/execExpr.c             | 11 ++++----
 src/backend/executor/nodeSubplan.c          | 28 +++++++++------------
 src/backend/replication/logical/relation.c  |  5 ++--
 src/backend/replication/logical/tablesync.c |  6 ++---
 src/backend/replication/pgoutput/pgoutput.c |  8 +++---
 5 files changed, 25 insertions(+), 33 deletions(-)

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index 2c62b0c9c84..e73261ec445 100644
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -216,7 +216,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	ExprState  *state;
 	ExprEvalStep scratch = {0};
 	List	   *adjust_jumps = NIL;
-	ListCell   *lc;
+	Expr	   *node;
+	int			jump;
 
 	/* short-circuit (here and in ExecQual) for empty restriction list */
 	if (qual == NIL)
@@ -250,10 +251,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	scratch.resvalue = &state->resvalue;
 	scratch.resnull = &state->resnull;
 
-	foreach(lc, qual)
+	for_each_ptr(node, qual)
 	{
-		Expr	   *node = (Expr *) lfirst(lc);
-
 		/* first evaluate expression */
 		ExecInitExprRec(node, state, &state->resvalue, &state->resnull);
 
@@ -265,9 +264,9 @@ ExecInitQual(List *qual, PlanState *parent)
 	}
 
 	/* adjust jump targets */
-	foreach(lc, adjust_jumps)
+	for_each_int(jump, adjust_jumps)
 	{
-		ExprEvalStep *as = &state->steps[lfirst_int(lc)];
+		ExprEvalStep *as = &state->steps[jump];
 
 		Assert(as->opcode == EEOP_QUAL);
 		Assert(as->d.qualexpr.jumpdone == -1);
diff --git a/src/backend/executor/nodeSubplan.c b/src/backend/executor/nodeSubplan.c
index c136f75ac24..8f88f289269 100644
--- a/src/backend/executor/nodeSubplan.c
+++ b/src/backend/executor/nodeSubplan.c
@@ -307,7 +307,7 @@ ExecScanSubPlan(SubPlanState *node,
 		Datum		rowresult;
 		bool		rownull;
 		int			col;
-		ListCell   *plst;
+		int			paramid;
 
 		if (subLinkType == EXISTS_SUBLINK)
 		{
@@ -367,9 +367,8 @@ ExecScanSubPlan(SubPlanState *node,
 			 * Now set all the setParam params from the columns of the tuple
 			 */
 			col = 1;
-			foreach(plst, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(plst);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -412,9 +411,8 @@ ExecScanSubPlan(SubPlanState *node,
 		 * combining expression.
 		 */
 		col = 1;
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -480,10 +478,11 @@ ExecScanSubPlan(SubPlanState *node,
 		}
 		else if (subLinkType == MULTIEXPR_SUBLINK)
 		{
+			int			paramid;
+
 			/* We don't care about function result, but set the setParams */
-			foreach(l, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(l);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -604,16 +603,15 @@ buildSubPlanHash(SubPlanState *node, ExprContext *econtext)
 		 slot = ExecProcNode(planstate))
 	{
 		int			col = 1;
-		ListCell   *plst;
 		bool		isnew;
+		int			paramid;
 
 		/*
 		 * Load up the Params representing the raw sub-select outputs, then
 		 * form the projection tuple to store in the hashtable.
 		 */
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
@@ -880,11 +878,10 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 	if (subplan->setParam != NIL && subplan->parParam == NIL &&
 		subplan->subLinkType != CTE_SUBLINK)
 	{
-		ListCell   *lst;
+		int			paramid;
 
-		foreach(lst, subplan->setParam)
+		for_each_int(paramid, subplan->setParam)
 		{
-			int			paramid = lfirst_int(lst);
 			ParamExecData *prm = &(estate->es_param_exec_vals[paramid]);
 
 			prm->execPlan = sstate;
@@ -906,7 +903,7 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		List	   *oplist,
 				   *lefttlist,
 				   *righttlist;
-		ListCell   *l;
+		OpExpr	   *opexpr;
 
 		/* We need a memory context to hold the hash table(s) */
 		sstate->hashtablecxt =
@@ -966,9 +963,8 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		cross_eq_funcoids = (Oid *) palloc(ncols * sizeof(Oid));
 
 		i = 1;
-		foreach(l, oplist)
+		for_each_node(OpExpr, opexpr, oplist)
 		{
-			OpExpr	   *opexpr = lfirst_node(OpExpr, l);
 			Expr	   *expr;
 			TargetEntry *tle;
 			Oid			rhs_eq_oper;
diff --git a/src/backend/replication/logical/relation.c b/src/backend/replication/logical/relation.c
index d62eefed138..f8faceb57ce 100644
--- a/src/backend/replication/logical/relation.c
+++ b/src/backend/replication/logical/relation.c
@@ -747,11 +747,10 @@ static Oid
 FindUsableIndexForReplicaIdentityFull(Relation localrel, AttrMap *attrmap)
 {
 	List	   *idxlist = RelationGetIndexList(localrel);
-	ListCell   *lc;
+	Oid			idxoid;
 
-	foreach(lc, idxlist)
+	for_each_oid(idxoid, idxlist)
 	{
-		Oid			idxoid = lfirst_oid(lc);
 		bool		isUsableIdx;
 		Relation	idxRel;
 		IndexInfo  *idxInfo;
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 37a0abe2f4d..c82d05a642e 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	ListCell   *lc;
+	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,10 +453,8 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	foreach(lc, table_states_not_ready)
+	for_each_ptr(rstate, table_states_not_ready)
 	{
-		SubscriptionRelState *rstate = (SubscriptionRelState *) lfirst(lc);
-
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
 			/*
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index c1c66848f36..ec6df59bedc 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -2229,7 +2229,7 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 {
 	HASH_SEQ_STATUS hash_seq;
 	RelationSyncEntry *entry;
-	ListCell   *lc;
+	TransactionId streamed_txn;
 
 	Assert(RelationSyncCache != NULL);
 
@@ -2242,15 +2242,15 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 		 * corresponding schema and we don't need to send it unless there is
 		 * any invalidation for that relation.
 		 */
-		foreach(lc, entry->streamed_txns)
+		for_each_xid(streamed_txn, entry->streamed_txns)
 		{
-			if (xid == lfirst_xid(lc))
+			if (xid == streamed_txn)
 			{
 				if (is_commit)
 					entry->schema_sent = true;
 
 				entry->streamed_txns =
-					foreach_delete_current(entry->streamed_txns, lc);
+					foreach_delete_current(entry->streamed_txns, streamed_txn);
 				break;
 			}
 		}
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-25 10:39  Jelte Fennema <[email protected]>
  parent: Jelte Fennema <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Jelte Fennema @ 2023-10-25 10:39 UTC (permalink / raw)
  To: David Rowley <[email protected]>; +Cc: Nathan Bossart <[email protected]>; pgsql-hackers

Attached is a slightly updated version, with a bit simpler
implementation of foreach_delete_current.
Instead of decrementing i and then adding 1 to it when indexing the
list, it now indexes the list using a postfix decrement.


Attachments:

  [application/x-patch] v4-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch (7.3K, ../../CAGECzQRA+sZH94TsN_Rewi3Fk6EeQueCUvuep+ngdXSqnvZ=kw@mail.gmail.com/2-v4-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch)
  download | inline diff:
From 9073777a6d82e2b2db8b1ed9aef200550234d89a Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 17:13:20 +0200
Subject: [PATCH v4 2/2] Use new for_each_xyz macros in a few places

This starts using each of the newly added for_each_xyz macros in at
least one place. This shows how they improve readability by reducing the
amount of boilerplate code.
---
 src/backend/executor/execExpr.c             | 11 ++++----
 src/backend/executor/nodeSubplan.c          | 28 +++++++++------------
 src/backend/replication/logical/relation.c  |  5 ++--
 src/backend/replication/logical/tablesync.c |  6 ++---
 src/backend/replication/pgoutput/pgoutput.c |  8 +++---
 5 files changed, 25 insertions(+), 33 deletions(-)

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index 2c62b0c9c84..e73261ec445 100644
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -216,7 +216,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	ExprState  *state;
 	ExprEvalStep scratch = {0};
 	List	   *adjust_jumps = NIL;
-	ListCell   *lc;
+	Expr	   *node;
+	int			jump;
 
 	/* short-circuit (here and in ExecQual) for empty restriction list */
 	if (qual == NIL)
@@ -250,10 +251,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	scratch.resvalue = &state->resvalue;
 	scratch.resnull = &state->resnull;
 
-	foreach(lc, qual)
+	for_each_ptr(node, qual)
 	{
-		Expr	   *node = (Expr *) lfirst(lc);
-
 		/* first evaluate expression */
 		ExecInitExprRec(node, state, &state->resvalue, &state->resnull);
 
@@ -265,9 +264,9 @@ ExecInitQual(List *qual, PlanState *parent)
 	}
 
 	/* adjust jump targets */
-	foreach(lc, adjust_jumps)
+	for_each_int(jump, adjust_jumps)
 	{
-		ExprEvalStep *as = &state->steps[lfirst_int(lc)];
+		ExprEvalStep *as = &state->steps[jump];
 
 		Assert(as->opcode == EEOP_QUAL);
 		Assert(as->d.qualexpr.jumpdone == -1);
diff --git a/src/backend/executor/nodeSubplan.c b/src/backend/executor/nodeSubplan.c
index c136f75ac24..8f88f289269 100644
--- a/src/backend/executor/nodeSubplan.c
+++ b/src/backend/executor/nodeSubplan.c
@@ -307,7 +307,7 @@ ExecScanSubPlan(SubPlanState *node,
 		Datum		rowresult;
 		bool		rownull;
 		int			col;
-		ListCell   *plst;
+		int			paramid;
 
 		if (subLinkType == EXISTS_SUBLINK)
 		{
@@ -367,9 +367,8 @@ ExecScanSubPlan(SubPlanState *node,
 			 * Now set all the setParam params from the columns of the tuple
 			 */
 			col = 1;
-			foreach(plst, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(plst);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -412,9 +411,8 @@ ExecScanSubPlan(SubPlanState *node,
 		 * combining expression.
 		 */
 		col = 1;
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -480,10 +478,11 @@ ExecScanSubPlan(SubPlanState *node,
 		}
 		else if (subLinkType == MULTIEXPR_SUBLINK)
 		{
+			int			paramid;
+
 			/* We don't care about function result, but set the setParams */
-			foreach(l, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(l);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -604,16 +603,15 @@ buildSubPlanHash(SubPlanState *node, ExprContext *econtext)
 		 slot = ExecProcNode(planstate))
 	{
 		int			col = 1;
-		ListCell   *plst;
 		bool		isnew;
+		int			paramid;
 
 		/*
 		 * Load up the Params representing the raw sub-select outputs, then
 		 * form the projection tuple to store in the hashtable.
 		 */
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
@@ -880,11 +878,10 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 	if (subplan->setParam != NIL && subplan->parParam == NIL &&
 		subplan->subLinkType != CTE_SUBLINK)
 	{
-		ListCell   *lst;
+		int			paramid;
 
-		foreach(lst, subplan->setParam)
+		for_each_int(paramid, subplan->setParam)
 		{
-			int			paramid = lfirst_int(lst);
 			ParamExecData *prm = &(estate->es_param_exec_vals[paramid]);
 
 			prm->execPlan = sstate;
@@ -906,7 +903,7 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		List	   *oplist,
 				   *lefttlist,
 				   *righttlist;
-		ListCell   *l;
+		OpExpr	   *opexpr;
 
 		/* We need a memory context to hold the hash table(s) */
 		sstate->hashtablecxt =
@@ -966,9 +963,8 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		cross_eq_funcoids = (Oid *) palloc(ncols * sizeof(Oid));
 
 		i = 1;
-		foreach(l, oplist)
+		for_each_node(OpExpr, opexpr, oplist)
 		{
-			OpExpr	   *opexpr = lfirst_node(OpExpr, l);
 			Expr	   *expr;
 			TargetEntry *tle;
 			Oid			rhs_eq_oper;
diff --git a/src/backend/replication/logical/relation.c b/src/backend/replication/logical/relation.c
index d62eefed138..f8faceb57ce 100644
--- a/src/backend/replication/logical/relation.c
+++ b/src/backend/replication/logical/relation.c
@@ -747,11 +747,10 @@ static Oid
 FindUsableIndexForReplicaIdentityFull(Relation localrel, AttrMap *attrmap)
 {
 	List	   *idxlist = RelationGetIndexList(localrel);
-	ListCell   *lc;
+	Oid			idxoid;
 
-	foreach(lc, idxlist)
+	for_each_oid(idxoid, idxlist)
 	{
-		Oid			idxoid = lfirst_oid(lc);
 		bool		isUsableIdx;
 		Relation	idxRel;
 		IndexInfo  *idxInfo;
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 37a0abe2f4d..c82d05a642e 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	ListCell   *lc;
+	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,10 +453,8 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	foreach(lc, table_states_not_ready)
+	for_each_ptr(rstate, table_states_not_ready)
 	{
-		SubscriptionRelState *rstate = (SubscriptionRelState *) lfirst(lc);
-
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
 			/*
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index c1c66848f36..ec6df59bedc 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -2229,7 +2229,7 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 {
 	HASH_SEQ_STATUS hash_seq;
 	RelationSyncEntry *entry;
-	ListCell   *lc;
+	TransactionId streamed_txn;
 
 	Assert(RelationSyncCache != NULL);
 
@@ -2242,15 +2242,15 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 		 * corresponding schema and we don't need to send it unless there is
 		 * any invalidation for that relation.
 		 */
-		foreach(lc, entry->streamed_txns)
+		for_each_xid(streamed_txn, entry->streamed_txns)
 		{
-			if (xid == lfirst_xid(lc))
+			if (xid == streamed_txn)
 			{
 				if (is_commit)
 					entry->schema_sent = true;
 
 				entry->streamed_txns =
-					foreach_delete_current(entry->streamed_txns, lc);
+					foreach_delete_current(entry->streamed_txns, streamed_txn);
 				break;
 			}
 		}
-- 
2.34.1



  [application/x-patch] v4-0001-Add-macros-for-looping-through-a-list-without-nee.patch (6.0K, ../../CAGECzQRA+sZH94TsN_Rewi3Fk6EeQueCUvuep+ngdXSqnvZ=kw@mail.gmail.com/3-v4-0001-Add-macros-for-looping-through-a-list-without-nee.patch)
  download | inline diff:
From 4f5bbe8865a420d635d0d4357388d1c3632a9821 Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 16:46:08 +0200
Subject: [PATCH v4 1/2] Add macros for looping through a list without needing
 a ListCell

Many usages of the foreach macro only use the ListCell variable to get
its contents. This adds macros that simplify iteration code for that
common use case.
---
 src/include/nodes/pg_list.h | 111 ++++++++++++++++++++++++++++++++++--
 1 file changed, 105 insertions(+), 6 deletions(-)

diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index 529a382d284..e09116e8fdb 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -383,13 +383,14 @@ lnext(const List *l, const ListCell *c)
  *	  delete the current list element from the List associated with a
  *	  surrounding foreach() loop, returning the new List pointer.
  *
- * This is equivalent to list_delete_cell(), but it also adjusts the foreach
- * loop's state so that no list elements will be missed.  Do not delete
- * elements from an active foreach loop's list in any other way!
+ * This is similar to list_delete_cell(), but it also works when using
+ * for_each_xyz macros that don't require passing in a "ListCell *".
+ * Furthermore it adjusts the foreach loop's state so that no list elements
+ * will be missed. Do not delete elements from an active foreach loop's list in
+ * any other way!
  */
-#define foreach_delete_current(lst, cell)	\
-	(cell##__state.i--, \
-	 (List *) (cell##__state.l = list_delete_cell(lst, cell)))
+#define foreach_delete_current(lst, var)	\
+	((List *) (var##__state.l = list_delete_cell(lst, &var##__state.l->elements[var##__state.i--])))
 
 /*
  * foreach_current_index -
@@ -452,6 +453,104 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 	return r;
 }
 
+/*
+ * for_each_ptr -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *", just a declared pointer variable to store the
+ *	  current pointer int;
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_ptr(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_int -
+ *	  a convenience macro which loops through a list of ints without needing a
+ *	  "ListCell *", just a declared int variable to store the current int in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_int(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_int(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_oid -
+ *	  a convenience macro which loops through an oid list without needing a
+ *	  "ListCell *", just a declared Oid variable to store the current oid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_oid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_oid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_xid -
+ *	  a convenience macro which loops through an xid list without needing a
+ *	  "ListCell *", just a declared TransactionId variable to store the current
+ *	  xid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_xid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_xid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_node -
+ *	  a convenience macro which loops through a node list without needing a
+ *	  "ListCell *", just a declared pointer variable to store the pointer of
+ *	  the current node in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_node(type, var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_node(type, &var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
 /*
  * forboth -
  *	  a convenience macro for advancing through two linked lists

base-commit: 8f0fd47fa33720dd09ad0ae74a8a583b9780e328
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 53+ messages in thread

* Re: Add new for_each macros for iterating over a List that do not require ListCell pointer
@ 2023-10-25 12:35  Jelte Fennema <[email protected]>
  parent: Alvaro Herrera <[email protected]>
  0 siblings, 0 replies; 53+ messages in thread

From: Jelte Fennema @ 2023-10-25 12:35 UTC (permalink / raw)
  To: Alvaro Herrera <[email protected]>; +Cc: pgsql-hackers

On Wed, 25 Oct 2023 at 13:52, Alvaro Herrera <[email protected]> wrote:
> Looking at for_each_ptr() I think it may be cleaner to follow
> palloc_object()'s precedent and make it foreach_object() instead (I have
> no love for the extra underscore, but I won't object to it either).  And
> like foreach_node, have it receive a type name to add a cast to.
>
> I'd imagine something like
>
>   SubscriptionRelState *rstate;
>
>   foreach_object(SubscriptionRelState *, rstate, table_states_not_ready)
>   {

Could you clarify why you think it may be cleaner? I don't see much
benefit to passing the type in there if all we use it for is adding a
cast. It seems like extra things to type for little benefit.
palloc_object uses the passed in type to not only do the cast, but
also to determine the size of the the allocation.

If foreach_object would allow us to remove the declaration further up
in the function I do see a benefit though.

I attached a new patchset which includes a 3rd patch that does this
(the other patches are equivalent to v4). I quite like that it moves
the type declaration to the loop itself, limiting its scope. But I'm
not fully convinced it's worth the hackiness of introducing a second
for loop that does a single iteration, just to be able to declare a
variable of a different type though. But I don't know another way of
achieving this. If this hack/trick is deemed acceptable, we can do the
same for the other newly introduced macros. The type would not even
need to be specified for oid/xid/int because it's already known to be
Oid/TransactionId/int respectively.


Attachments:

  [application/octet-stream] v5-0001-Add-macros-for-looping-through-a-list-without-nee.patch (6.0K, ../../CAGECzQTn-teFSHNu0en0D8kr9QohqQN9ebiNeTjXencn+uRXgg@mail.gmail.com/2-v5-0001-Add-macros-for-looping-through-a-list-without-nee.patch)
  download | inline diff:
From 4f5bbe8865a420d635d0d4357388d1c3632a9821 Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 16:46:08 +0200
Subject: [PATCH v5 1/3] Add macros for looping through a list without needing
 a ListCell

Many usages of the foreach macro only use the ListCell variable to get
its contents. This adds macros that simplify iteration code for that
common use case.
---
 src/include/nodes/pg_list.h | 111 ++++++++++++++++++++++++++++++++++--
 1 file changed, 105 insertions(+), 6 deletions(-)

diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index 529a382d284..e09116e8fdb 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -383,13 +383,14 @@ lnext(const List *l, const ListCell *c)
  *	  delete the current list element from the List associated with a
  *	  surrounding foreach() loop, returning the new List pointer.
  *
- * This is equivalent to list_delete_cell(), but it also adjusts the foreach
- * loop's state so that no list elements will be missed.  Do not delete
- * elements from an active foreach loop's list in any other way!
+ * This is similar to list_delete_cell(), but it also works when using
+ * for_each_xyz macros that don't require passing in a "ListCell *".
+ * Furthermore it adjusts the foreach loop's state so that no list elements
+ * will be missed. Do not delete elements from an active foreach loop's list in
+ * any other way!
  */
-#define foreach_delete_current(lst, cell)	\
-	(cell##__state.i--, \
-	 (List *) (cell##__state.l = list_delete_cell(lst, cell)))
+#define foreach_delete_current(lst, var)	\
+	((List *) (var##__state.l = list_delete_cell(lst, &var##__state.l->elements[var##__state.i--])))
 
 /*
  * foreach_current_index -
@@ -452,6 +453,104 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 	return r;
 }
 
+/*
+ * for_each_ptr -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *", just a declared pointer variable to store the
+ *	  current pointer int;
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_ptr(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_int -
+ *	  a convenience macro which loops through a list of ints without needing a
+ *	  "ListCell *", just a declared int variable to store the current int in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_int(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_int(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_oid -
+ *	  a convenience macro which loops through an oid list without needing a
+ *	  "ListCell *", just a declared Oid variable to store the current oid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_oid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_oid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * foreach_xid -
+ *	  a convenience macro which loops through an xid list without needing a
+ *	  "ListCell *", just a declared TransactionId variable to store the current
+ *	  xid in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_xid(var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_xid(&var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
+/*
+ * for_each_node -
+ *	  a convenience macro which loops through a node list without needing a
+ *	  "ListCell *", just a declared pointer variable to store the pointer of
+ *	  the current node in.
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_node(type, var, lst) \
+	for (ForEachState var##__state = {(lst), 0}; \
+		 (var##__state.l != NIL && \
+		  var##__state.i < var##__state.l->length && \
+		  (var = lfirst_node(type, &var##__state.l->elements[var##__state.i]), true));\
+		 var##__state.i++)
+
 /*
  * forboth -
  *	  a convenience macro for advancing through two linked lists

base-commit: 8f0fd47fa33720dd09ad0ae74a8a583b9780e328
-- 
2.34.1



  [application/octet-stream] v5-0003-Introduce-for_each_object.patch (2.7K, ../../CAGECzQTn-teFSHNu0en0D8kr9QohqQN9ebiNeTjXencn+uRXgg@mail.gmail.com/3-v5-0003-Introduce-for_each_object.patch)
  download | inline diff:
From 6b21a0da96967bf94a7569e0a1c89bf13cf38868 Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Wed, 25 Oct 2023 14:20:07 +0200
Subject: [PATCH v5 3/3] Introduce for_each_object

Add new for_each_object helper that does not require var argument to be
declared before hand.
---
 src/backend/replication/logical/tablesync.c |  3 +--
 src/include/nodes/pg_list.h                 | 23 +++++++++++++++++++++
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index c82d05a642e..ed27181c2a9 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,6 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,7 +452,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	for_each_ptr(rstate, table_states_not_ready)
+	for_each_object(SubscriptionRelState, rstate, table_states_not_ready)
 	{
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
diff --git a/src/include/nodes/pg_list.h b/src/include/nodes/pg_list.h
index e09116e8fdb..0748d36536b 100644
--- a/src/include/nodes/pg_list.h
+++ b/src/include/nodes/pg_list.h
@@ -473,6 +473,29 @@ for_each_cell_setup(const List *lst, const ListCell *initcell)
 		  (var = lfirst(&var##__state.l->elements[var##__state.i]), true));\
 		 var##__state.i++)
 
+/*
+ * for_each_object -
+ *	  a convenience macro which loops through a list of pointers without
+ *	  needing a "ListCell *" only a type and variable name. It automatically
+ *	  declares a loop variable with the given name and type;
+ *
+ * Unlike with foreach() it's not possible to detect if an early "break"
+ * occured by checking the value of the loop variable at the end of the loop.
+ * If you need this, it's recommended to use foreach() instead or manually keep
+ * track of a break occured using a boolean flag variable called e.g. "found".
+ *
+ * The caveats for foreach() apply equally here.
+ */
+#define for_each_object(type, var, lst) \
+	for (type * var = NULL, *var##__outerloop = (type *) 1; \
+		 var##__outerloop; \
+		 var##__outerloop = NULL) \
+		for (ForEachState var##__state = {(lst), 0}; \
+			 (var##__state.l != NIL && \
+			  var##__state.i < var##__state.l->length && \
+			  (var = lfirst(&var##__state.l->elements[var##__state.i]), true));\
+			 var##__state.i++)
+
 /*
  * for_each_int -
  *	  a convenience macro which loops through a list of ints without needing a
-- 
2.34.1



  [application/octet-stream] v5-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch (7.3K, ../../CAGECzQTn-teFSHNu0en0D8kr9QohqQN9ebiNeTjXencn+uRXgg@mail.gmail.com/4-v5-0002-Use-new-for_each_xyz-macros-in-a-few-places.patch)
  download | inline diff:
From 9073777a6d82e2b2db8b1ed9aef200550234d89a Mon Sep 17 00:00:00 2001
From: Jelte Fennema-Nio <[email protected]>
Date: Tue, 24 Oct 2023 17:13:20 +0200
Subject: [PATCH v5 2/3] Use new for_each_xyz macros in a few places

This starts using each of the newly added for_each_xyz macros in at
least one place. This shows how they improve readability by reducing the
amount of boilerplate code.
---
 src/backend/executor/execExpr.c             | 11 ++++----
 src/backend/executor/nodeSubplan.c          | 28 +++++++++------------
 src/backend/replication/logical/relation.c  |  5 ++--
 src/backend/replication/logical/tablesync.c |  6 ++---
 src/backend/replication/pgoutput/pgoutput.c |  8 +++---
 5 files changed, 25 insertions(+), 33 deletions(-)

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index 2c62b0c9c84..e73261ec445 100644
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -216,7 +216,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	ExprState  *state;
 	ExprEvalStep scratch = {0};
 	List	   *adjust_jumps = NIL;
-	ListCell   *lc;
+	Expr	   *node;
+	int			jump;
 
 	/* short-circuit (here and in ExecQual) for empty restriction list */
 	if (qual == NIL)
@@ -250,10 +251,8 @@ ExecInitQual(List *qual, PlanState *parent)
 	scratch.resvalue = &state->resvalue;
 	scratch.resnull = &state->resnull;
 
-	foreach(lc, qual)
+	for_each_ptr(node, qual)
 	{
-		Expr	   *node = (Expr *) lfirst(lc);
-
 		/* first evaluate expression */
 		ExecInitExprRec(node, state, &state->resvalue, &state->resnull);
 
@@ -265,9 +264,9 @@ ExecInitQual(List *qual, PlanState *parent)
 	}
 
 	/* adjust jump targets */
-	foreach(lc, adjust_jumps)
+	for_each_int(jump, adjust_jumps)
 	{
-		ExprEvalStep *as = &state->steps[lfirst_int(lc)];
+		ExprEvalStep *as = &state->steps[jump];
 
 		Assert(as->opcode == EEOP_QUAL);
 		Assert(as->d.qualexpr.jumpdone == -1);
diff --git a/src/backend/executor/nodeSubplan.c b/src/backend/executor/nodeSubplan.c
index c136f75ac24..8f88f289269 100644
--- a/src/backend/executor/nodeSubplan.c
+++ b/src/backend/executor/nodeSubplan.c
@@ -307,7 +307,7 @@ ExecScanSubPlan(SubPlanState *node,
 		Datum		rowresult;
 		bool		rownull;
 		int			col;
-		ListCell   *plst;
+		int			paramid;
 
 		if (subLinkType == EXISTS_SUBLINK)
 		{
@@ -367,9 +367,8 @@ ExecScanSubPlan(SubPlanState *node,
 			 * Now set all the setParam params from the columns of the tuple
 			 */
 			col = 1;
-			foreach(plst, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(plst);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -412,9 +411,8 @@ ExecScanSubPlan(SubPlanState *node,
 		 * combining expression.
 		 */
 		col = 1;
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -480,10 +478,11 @@ ExecScanSubPlan(SubPlanState *node,
 		}
 		else if (subLinkType == MULTIEXPR_SUBLINK)
 		{
+			int			paramid;
+
 			/* We don't care about function result, but set the setParams */
-			foreach(l, subplan->setParam)
+			for_each_int(paramid, subplan->setParam)
 			{
-				int			paramid = lfirst_int(l);
 				ParamExecData *prmdata;
 
 				prmdata = &(econtext->ecxt_param_exec_vals[paramid]);
@@ -604,16 +603,15 @@ buildSubPlanHash(SubPlanState *node, ExprContext *econtext)
 		 slot = ExecProcNode(planstate))
 	{
 		int			col = 1;
-		ListCell   *plst;
 		bool		isnew;
+		int			paramid;
 
 		/*
 		 * Load up the Params representing the raw sub-select outputs, then
 		 * form the projection tuple to store in the hashtable.
 		 */
-		foreach(plst, subplan->paramIds)
+		for_each_int(paramid, subplan->paramIds)
 		{
-			int			paramid = lfirst_int(plst);
 			ParamExecData *prmdata;
 
 			prmdata = &(innerecontext->ecxt_param_exec_vals[paramid]);
@@ -880,11 +878,10 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 	if (subplan->setParam != NIL && subplan->parParam == NIL &&
 		subplan->subLinkType != CTE_SUBLINK)
 	{
-		ListCell   *lst;
+		int			paramid;
 
-		foreach(lst, subplan->setParam)
+		for_each_int(paramid, subplan->setParam)
 		{
-			int			paramid = lfirst_int(lst);
 			ParamExecData *prm = &(estate->es_param_exec_vals[paramid]);
 
 			prm->execPlan = sstate;
@@ -906,7 +903,7 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		List	   *oplist,
 				   *lefttlist,
 				   *righttlist;
-		ListCell   *l;
+		OpExpr	   *opexpr;
 
 		/* We need a memory context to hold the hash table(s) */
 		sstate->hashtablecxt =
@@ -966,9 +963,8 @@ ExecInitSubPlan(SubPlan *subplan, PlanState *parent)
 		cross_eq_funcoids = (Oid *) palloc(ncols * sizeof(Oid));
 
 		i = 1;
-		foreach(l, oplist)
+		for_each_node(OpExpr, opexpr, oplist)
 		{
-			OpExpr	   *opexpr = lfirst_node(OpExpr, l);
 			Expr	   *expr;
 			TargetEntry *tle;
 			Oid			rhs_eq_oper;
diff --git a/src/backend/replication/logical/relation.c b/src/backend/replication/logical/relation.c
index d62eefed138..f8faceb57ce 100644
--- a/src/backend/replication/logical/relation.c
+++ b/src/backend/replication/logical/relation.c
@@ -747,11 +747,10 @@ static Oid
 FindUsableIndexForReplicaIdentityFull(Relation localrel, AttrMap *attrmap)
 {
 	List	   *idxlist = RelationGetIndexList(localrel);
-	ListCell   *lc;
+	Oid			idxoid;
 
-	foreach(lc, idxlist)
+	for_each_oid(idxoid, idxlist)
 	{
-		Oid			idxoid = lfirst_oid(lc);
 		bool		isUsableIdx;
 		Relation	idxRel;
 		IndexInfo  *idxInfo;
diff --git a/src/backend/replication/logical/tablesync.c b/src/backend/replication/logical/tablesync.c
index 37a0abe2f4d..c82d05a642e 100644
--- a/src/backend/replication/logical/tablesync.c
+++ b/src/backend/replication/logical/tablesync.c
@@ -416,7 +416,7 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 		TimestampTz last_start_time;
 	};
 	static HTAB *last_start_times = NULL;
-	ListCell   *lc;
+	SubscriptionRelState *rstate;
 	bool		started_tx = false;
 	bool		should_exit = false;
 
@@ -453,10 +453,8 @@ process_syncing_tables_for_apply(XLogRecPtr current_lsn)
 	/*
 	 * Process all tables that are being synchronized.
 	 */
-	foreach(lc, table_states_not_ready)
+	for_each_ptr(rstate, table_states_not_ready)
 	{
-		SubscriptionRelState *rstate = (SubscriptionRelState *) lfirst(lc);
-
 		if (rstate->state == SUBREL_STATE_SYNCDONE)
 		{
 			/*
diff --git a/src/backend/replication/pgoutput/pgoutput.c b/src/backend/replication/pgoutput/pgoutput.c
index c1c66848f36..ec6df59bedc 100644
--- a/src/backend/replication/pgoutput/pgoutput.c
+++ b/src/backend/replication/pgoutput/pgoutput.c
@@ -2229,7 +2229,7 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 {
 	HASH_SEQ_STATUS hash_seq;
 	RelationSyncEntry *entry;
-	ListCell   *lc;
+	TransactionId streamed_txn;
 
 	Assert(RelationSyncCache != NULL);
 
@@ -2242,15 +2242,15 @@ cleanup_rel_sync_cache(TransactionId xid, bool is_commit)
 		 * corresponding schema and we don't need to send it unless there is
 		 * any invalidation for that relation.
 		 */
-		foreach(lc, entry->streamed_txns)
+		for_each_xid(streamed_txn, entry->streamed_txns)
 		{
-			if (xid == lfirst_xid(lc))
+			if (xid == streamed_txn)
 			{
 				if (is_commit)
 					entry->schema_sent = true;
 
 				entry->streamed_txns =
-					foreach_delete_current(entry->streamed_txns, lc);
+					foreach_delete_current(entry->streamed_txns, streamed_txn);
 				break;
 			}
 		}
-- 
2.34.1



^ permalink  raw  reply  [nested|flat] 53+ messages in thread

end of thread, other threads:[~2023-10-25 12:35 UTC | newest]

Thread overview: 53+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 03:25 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2021-02-04 18:43 [PATCH] hint squash commit Bruce Momjian <[email protected]>
2023-10-24 16:03 Add new for_each macros for iterating over a List that do not require ListCell pointer Jelte Fennema <[email protected]>
2023-10-24 16:47 ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Nathan Bossart <[email protected]>
2023-10-24 16:58   ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Jelte Fennema <[email protected]>
2023-10-24 21:20     ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Nathan Bossart <[email protected]>
2023-10-25 02:55     ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer David Rowley <[email protected]>
2023-10-25 10:05       ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Jelte Fennema <[email protected]>
2023-10-25 10:39         ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Jelte Fennema <[email protected]>
2023-10-25 08:51 ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Alvaro Herrera <[email protected]>
2023-10-25 12:35   ` Re: Add new for_each macros for iterating over a List that do not require ListCell pointer Jelte Fennema <[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