0003-Apply-wal_sync_method-to-pg_control-file.patch

From c9dcb77263a7f86b187231a83083bfbe3a0f30b4 Mon Sep 17 00:00:00 2001
From: Thomas Munro <thomas.munro@gmail.com>
Date: Tue, 31 Jan 2023 21:08:12 +1300
Subject: [PATCH 3/3] Apply wal_sync_method to pg_control file.

When writing out the control file frequently on a standby, we can go a
little faster on some systems by using fdatasync() instead of fsync(),
with no loss of safety.  We already respected the special
"wal_sync_method=fsync_writethrough" via a circuitous route, but we
might as well support the full range of methods.
---
 src/backend/access/transam/xlog.c      | 19 ++++++++-
 src/common/controldata_utils.c         | 59 ++++++++++++++++++++------
 src/include/common/controldata_utils.h |  7 ++-
 3 files changed, 70 insertions(+), 15 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 9b9c3294e8..27c1cb8273 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -4185,7 +4185,24 @@ ReadControlFile(void)
 static void
 UpdateControlFile(void)
 {
-	update_controlfile(DataDir, ControlFile, true);
+	int			sync_op;
+
+	switch (sync_method)
+	{
+		case SYNC_METHOD_FDATASYNC:
+			sync_op = UPDATE_CONTROLFILE_FDATASYNC;
+			break;
+		case SYNC_METHOD_OPEN:
+			sync_op = UPDATE_CONTROLFILE_O_SYNC;
+			break;
+		case SYNC_METHOD_OPEN_DSYNC:
+			sync_op = UPDATE_CONTROLFILE_O_DSYNC;
+			break;
+		default:
+			sync_op = UPDATE_CONTROLFILE_FSYNC;
+	}
+
+	update_controlfile(DataDir, ControlFile, sync_op);
 }
 
 /*
diff --git a/src/common/controldata_utils.c b/src/common/controldata_utils.c
index 45b43ae546..3fc2a00d44 100644
--- a/src/common/controldata_utils.c
+++ b/src/common/controldata_utils.c
@@ -239,18 +239,39 @@ get_controlfile(const char *DataDir, bool *crc_ok_p)
  * update_controlfile()
  *
  * Update controlfile values with the contents given by caller.  The
- * contents to write are included in "ControlFile". "do_sync" can be
- * optionally used to flush the updated control file.  Note that it is up
- * to the caller to properly lock ControlFileLock when calling this
- * routine in the backend.
+ * contents to write are included in "ControlFile".  "sync_op" can be set
+ * to 0 or a synchronization method UPDATE_CONTROLFILE_XXX.  In frontend code,
+ * only 0 and non-zero (fsync) are meaningful.
+ * Note that it is up to the caller to properly lock ControlFileLock when
+ * calling this routine in the backend.
  */
 void
 update_controlfile(const char *DataDir,
-				   ControlFileData *ControlFile, bool do_sync)
+				   ControlFileData *ControlFile,
+				   int sync_op)
 {
 	int			fd;
 	char		buffer[PG_CONTROL_FILE_SIZE];
 	char		ControlFilePath[MAXPGPATH];
+	int			open_flag;
+
+	switch (sync_op)
+	{
+#ifndef FRONTEND
+#ifdef O_SYNC
+		case UPDATE_CONTROLFILE_O_SYNC:
+			open_flag = O_SYNC;
+			break;
+#endif
+#ifdef O_DSYNC
+		case UPDATE_CONTROLFILE_O_DSYNC:
+			open_flag = O_DSYNC;
+			break;
+#endif
+#endif
+		default:
+			open_flag = 0;
+	}
 
 	/* Update timestamp  */
 	ControlFile->time = (pg_time_t) time(NULL);
@@ -278,13 +299,13 @@ update_controlfile(const char *DataDir,
 	 * All errors issue a PANIC, so no need to use OpenTransientFile() and to
 	 * worry about file descriptor leaks.
 	 */
-	if ((fd = BasicOpenFile(ControlFilePath, O_RDWR | PG_BINARY)) < 0)
+	if ((fd = BasicOpenFile(ControlFilePath, O_RDWR | PG_BINARY | open_flag)) < 0)
 		ereport(PANIC,
 				(errcode_for_file_access(),
 				 errmsg("could not open file \"%s\": %m",
 						ControlFilePath)));
 #else
-	if ((fd = open(ControlFilePath, O_WRONLY | PG_BINARY,
+	if ((fd = open(ControlFilePath, O_WRONLY | PG_BINARY | open_flag,
 				   pg_file_create_mode)) == -1)
 		pg_fatal("could not open file \"%s\": %m", ControlFilePath);
 #endif
@@ -331,17 +352,29 @@ update_controlfile(const char *DataDir,
 	pgstat_report_wait_end();
 #endif
 
-	if (do_sync)
+	if (sync_op != 0)
 	{
 #ifndef FRONTEND
 		pgstat_report_wait_start(WAIT_EVENT_CONTROL_FILE_SYNC_UPDATE);
-		if (pg_fsync(fd) != 0)
-			ereport(PANIC,
-					(errcode_for_file_access(),
-					 errmsg("could not fsync file \"%s\": %m",
-							ControlFilePath)));
+		if (sync_op == UPDATE_CONTROLFILE_FDATASYNC)
+		{
+			if (fdatasync(fd) != 0)
+				ereport(PANIC,
+						(errcode_for_file_access(),
+						 errmsg("could not fdatasync file \"%s\": %m",
+								ControlFilePath)));
+		}
+		else if (sync_op == UPDATE_CONTROLFILE_FSYNC)
+		{
+			if (pg_fsync(fd) != 0)
+				ereport(PANIC,
+						(errcode_for_file_access(),
+						 errmsg("could not fdatasync file \"%s\": %m",
+								ControlFilePath)));
+		}
 		pgstat_report_wait_end();
 #else
+		/* In frontend code, non-zero sync_op gets you plain fsync(). */
 		if (fsync(fd) != 0)
 			pg_fatal("could not fsync file \"%s\": %m", ControlFilePath);
 #endif
diff --git a/src/include/common/controldata_utils.h b/src/include/common/controldata_utils.h
index 06825cad85..de6de5571e 100644
--- a/src/include/common/controldata_utils.h
+++ b/src/include/common/controldata_utils.h
@@ -12,9 +12,14 @@
 
 #include "catalog/pg_control.h"
 
+#define UPDATE_CONTROLFILE_FSYNC 1
+#define UPDATE_CONTROLFILE_FDATASYNC 2
+#define UPDATE_CONTROLFILE_O_SYNC 3
+#define UPDATE_CONTROLFILE_O_DSYNC 4
+
 extern ControlFileData *get_controlfile(const char *DataDir, bool *crc_ok_p);
 extern void update_controlfile(const char *DataDir,
-							   ControlFileData *ControlFile, bool do_sync);
+							   ControlFileData *ControlFile, int sync_op);
 extern int lock_controlfile(int fd, bool exclusive);
 #ifdef WIN32
 extern int unlock_controlfile(int fd);
-- 
2.35.1