v3-0003-Handle-rewind-failure-when-a-timeline-ends-with-an-overwritten-contrecord.patch

From a189e0d922c19a23739a4508b0e722c7459e872d Mon Sep 17 00:00:00 2001
From: Alena Vinter <dlaaren8@gmail.com>
Date: Wed, 10 Sep 2025 14:06:09 +0700
Subject: [PATCH 3/3] Handle rewind failure when a timeline ends with an
 overwritten contrecord.

When a common timeline ends with an overwritten contrecord, the
divergence point may not point to the start of a valid WAL record on the
target, causing errors and making rewind impossible.

To handle this case, when the target timeline is unfinished, we look for
a checkpoint preceding the divergence point starting from the last
checkpoint on the target rather than from the divergence point itself.
This ensures we always begin from a known-valid position in WAL.
---
 src/bin/pg_rewind/parsexlog.c | 25 +++++++++++++------------
 src/bin/pg_rewind/pg_rewind.c | 18 ++++++++++++++++--
 src/bin/pg_rewind/pg_rewind.h |  7 ++++---
 3 files changed, 33 insertions(+), 17 deletions(-)

diff --git a/src/bin/pg_rewind/parsexlog.c b/src/bin/pg_rewind/parsexlog.c
index 8f4b282c6b1..7d01fefb1d5 100644
--- a/src/bin/pg_rewind/parsexlog.c
+++ b/src/bin/pg_rewind/parsexlog.c
@@ -165,9 +165,10 @@ readOneRecord(const char *datadir, XLogRecPtr ptr, int tliIndex,
  * Find the previous checkpoint preceding given WAL location.
  */
 void
-findLastCheckpoint(const char *datadir, XLogRecPtr forkptr, int tliIndex,
-				   XLogRecPtr *lastchkptrec, TimeLineID *lastchkpttli,
-				   XLogRecPtr *lastchkptredo, const char *restoreCommand)
+findLastCheckpoint(const char *datadir, XLogRecPtr startptr, XLogRecPtr forkptr,
+				   int tliIndex, XLogRecPtr *lastchkptrec,
+				   TimeLineID *lastchkpttli, XLogRecPtr *lastchkptredo,
+				   const char *restoreCommand)
 {
 	/* Walk backwards, starting from the given record */
 	XLogRecord *record;
@@ -179,17 +180,17 @@ findLastCheckpoint(const char *datadir, XLogRecPtr forkptr, int tliIndex,
 	TimeLineID	current_tli = 0;

 	/*
-	 * The given fork pointer points to the end of the last common record,
-	 * which is not necessarily the beginning of the next record, if the
-	 * previous record happens to end at a page boundary. Skip over the page
-	 * header in that case to find the next record.
+	 * The given start pointer may point to a page boundary if the startptr is
+	 * the end of the last common record which is not necessarily the beginning
+	 * of the next record. Skip over the page header in that case to find the
+	 * next record.
 	 */
-	if (forkptr % XLOG_BLCKSZ == 0)
+	if (startptr % XLOG_BLCKSZ == 0)
 	{
-		if (XLogSegmentOffset(forkptr, WalSegSz) == 0)
-			forkptr += SizeOfXLogLongPHD;
+		if (XLogSegmentOffset(startptr, WalSegSz) == 0)
+			startptr += SizeOfXLogLongPHD;
 		else
-			forkptr += SizeOfXLogShortPHD;
+			startptr += SizeOfXLogShortPHD;
 	}

 	private.tliIndex = tliIndex;
@@ -200,7 +201,7 @@ findLastCheckpoint(const char *datadir, XLogRecPtr forkptr, int tliIndex,
 	if (xlogreader == NULL)
 		pg_fatal("out of memory while allocating a WAL reading processor");

-	searchptr = forkptr;
+	searchptr = startptr;
 	for (;;)
 	{
 		uint8		info;
diff --git a/src/bin/pg_rewind/pg_rewind.c b/src/bin/pg_rewind/pg_rewind.c
index 0c68dd4235e..c2ecf72cacf 100644
--- a/src/bin/pg_rewind/pg_rewind.c
+++ b/src/bin/pg_rewind/pg_rewind.c
@@ -140,6 +140,7 @@ main(int argc, char **argv)
 	int			option_index;
 	int			c;
 	XLogRecPtr	divergerec;
+	XLogRecPtr	checkpoint_search_startrec;
 	int			lastcommontliIndex;
 	XLogRecPtr	chkptrec;
 	TimeLineID	chkpttli;
@@ -459,8 +460,21 @@ main(int argc, char **argv)
 	/* Initialize hashtable that tracks WAL files protected from removal */
 	keepwal_init();

-	findLastCheckpoint(datadir_target, divergerec, lastcommontliIndex,
-					   &chkptrec, &chkpttli, &chkptredo, restore_command);
+	/*
+	 * If the last common timeline is incomplete on the target, a divergence
+	 * point from the source's finished timeline may not exist in the target's
+	 * WAL. Therefore, start searching for a checkpoint preceding the divergence
+	 * point from the last checkpoint on the target server to find a safe common
+	 * point.
+	 */
+	if (targetHistory[lastcommontliIndex].end == InvalidXLogRecPtr)
+		checkpoint_search_startrec = ControlFile_target.checkPoint;
+	else
+		checkpoint_search_startrec = divergerec;
+
+	findLastCheckpoint(datadir_target, checkpoint_search_startrec, divergerec,
+					   lastcommontliIndex, &chkptrec, &chkpttli, &chkptredo,
+					   restore_command);
 	pg_log_info("rewinding from last common checkpoint at %X/%08X on timeline %u",
 				LSN_FORMAT_ARGS(chkptrec), chkpttli);

diff --git a/src/bin/pg_rewind/pg_rewind.h b/src/bin/pg_rewind/pg_rewind.h
index 9cea144d2b2..4879be1d1d4 100644
--- a/src/bin/pg_rewind/pg_rewind.h
+++ b/src/bin/pg_rewind/pg_rewind.h
@@ -35,9 +35,10 @@ extern uint64 fetch_done;
 extern void extractPageMap(const char *datadir, XLogRecPtr startpoint,
 						   int tliIndex, XLogRecPtr endpoint,
 						   const char *restoreCommand);
-extern void findLastCheckpoint(const char *datadir, XLogRecPtr forkptr,
-							   int tliIndex,
-							   XLogRecPtr *lastchkptrec, TimeLineID *lastchkpttli,
+extern void findLastCheckpoint(const char *datadir, XLogRecPtr startptr,
+							   XLogRecPtr forkptr, int tliIndex,
+							   XLogRecPtr *lastchkptrec,
+							   TimeLineID *lastchkpttli,
 							   XLogRecPtr *lastchkptredo,
 							   const char *restoreCommand);
 extern XLogRecPtr readOneRecord(const char *datadir, XLogRecPtr ptr,
--
2.51.0