avoid multiple hard links to same WAL file after a crash

From d3c633e19555dc0cf98207ad5e7c08ab9ce85dc0 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Fri, 8 Apr 2022 11:48:17 -0700
Subject: [PATCH v2 1/1] Remove durable_rename_excl().

durable_rename_excl() attempts to avoid overwriting any existing
files by using link() and unlink(), but it falls back to rename()
on some platforms (e.g., Windows), which offers no such ovewrite
protection.  Most callers use durable_rename_excl() just in case
there is an existing file, but in practice there shouldn't be one.
basic_archive uses it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  My testing demonstrated that
it was possible to end up with two links to the same file in pg_wal
after a crash just before unlink() during WAL recycling.
Specifically, the test produced links to the same file for the
current WAL file and the next one because the half-recycled WAL
file was re-recycled upon restarting.  This seems likely to lead to
WAL corruption.

This change removes durable_rename_excl() and replaces all existing
calls with durable_rename().  This removes the protection against
accidentally overwriting an existing file, but some platforms are
already living without it, and ordinarily there shouldn't be one.

Author: Nathan Bossart
Reviewed-by: Robert Haas
Discussion: https://postgr.es/m/20220407182954.GA1231544%40nathanxps13
---
 contrib/basic_archive/basic_archive.c |  5 ++-
 src/backend/access/transam/timeline.c | 14 +-----
 src/backend/access/transam/xlog.c     |  8 +---
 src/backend/storage/file/fd.c         | 63 ---------------------------
 src/include/pg_config_manual.h        |  7 ---
 src/include/storage/fd.h              |  1 -
 6 files changed, 7 insertions(+), 91 deletions(-)

diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index e7efbfb9c3..ed33854c57 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -281,9 +281,10 @@ basic_archive_file_internal(const char *file, const char *path)
 
 	/*
 	 * Sync the temporary file to disk and move it to its final destination.
-	 * This will fail if destination already exists.
+	 * Note that this will overwrite any existing file, but this is only
+	 * possible if someone else created the file since the stat() above.
 	 */
-	(void) durable_rename_excl(temp, destination, ERROR);
+	(void) durable_rename(temp, destination, ERROR);
 
 	ereport(DEBUG1,
 			(errmsg("archived \"%s\" via basic_archive", file)));
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index be21968293..128f754e87 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -519,12 +514,7 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index a7814d4019..d19215ab24 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3323,14 +3323,10 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 14b77f2861..88645ed83d 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -807,69 +807,6 @@ durable_unlink(const char *fname, int elevel)
 	return 0;
 }
 
-/*
- * durable_rename_excl -- rename a file in a durable manner.
- *
- * Similar to durable_rename(), except that this routine tries (but does not
- * guarantee) not to overwrite the target file.
- *
- * Note that a crash in an unfortunate moment can leave you with two links to
- * the target file.
- *
- * Log errors with the caller specified severity.
- *
- * On Windows, using a hard link followed by unlink() causes concurrency
- * issues, while a simple rename() does not cause that, so be careful when
- * changing the logic of this routine.
- *
- * Returns 0 if the operation succeeded, -1 otherwise. Note that errno is not
- * valid upon return.
- */
-int
-durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
-{
-	/*
-	 * Ensure that, if we crash directly after the rename/link, a file with
-	 * valid contents is moved into place.
-	 */
-	if (fsync_fname_ext(oldfile, false, false, elevel) != 0)
-		return -1;
-
-#ifdef HAVE_WORKING_LINK
-	if (link(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-	unlink(oldfile);
-#else
-	if (rename(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-#endif
-
-	/*
-	 * Make change persistent in case of an OS crash, both the new entry and
-	 * its parent directory need to be flushed.
-	 */
-	if (fsync_fname_ext(newfile, false, false, elevel) != 0)
-		return -1;
-
-	/* Same for parent directory */
-	if (fsync_parent_path(newfile, elevel) != 0)
-		return -1;
-
-	return 0;
-}
-
 /*
  * InitFileAccess --- initialize this module during backend startup
  *
diff --git a/src/include/pg_config_manual.h b/src/include/pg_config_manual.h
index 84ce5a4a5d..830804fdfb 100644
--- a/src/include/pg_config_manual.h
+++ b/src/include/pg_config_manual.h
@@ -163,13 +163,6 @@
 #define USE_BARRIER_SMGRRELEASE
 #endif
 
-/*
- * Define this if your operating system supports link()
- */
-#if !defined(WIN32) && !defined(__CYGWIN__)
-#define HAVE_WORKING_LINK 1
-#endif
-
 /*
  * USE_POSIX_FADVISE controls whether Postgres will attempt to use the
  * posix_fadvise() kernel call.  Usually the automatic configure tests are
diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h
index 69549b000f..2b4a8e0ffe 100644
--- a/src/include/storage/fd.h
+++ b/src/include/storage/fd.h
@@ -187,7 +187,6 @@ extern void fsync_fname(const char *fname, bool isdir);
 extern int	fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel);
 extern int	durable_rename(const char *oldfile, const char *newfile, int loglevel);
 extern int	durable_unlink(const char *fname, int loglevel);
-extern int	durable_rename_excl(const char *oldfile, const char *newfile, int loglevel);
 extern void SyncDataDirectory(void);
 extern int	data_sync_elevel(int elevel);
 
-- 
2.25.1

From d489c2bff029db6e07e5028788faf869c35f886b Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 11:56:50 -0700
Subject: [PATCH v3 1/2] Replace calls to durable_rename_excl() with
 durable_rename().

durable_rename_excl() attempts to avoid overwriting any existing
files by using link() and unlink(), but it falls back to rename()
on some platforms (e.g., Windows), which offers no such overwrite
protection.  Most callers use durable_rename_excl() just in case
there is an existing file, but in practice there shouldn't be one.
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  My testing demonstrated that
it was possible to end up with two links to the same file in pg_wal
after a crash just before unlink() during WAL recycling.
Specifically, the test produced links to the same file for the
current WAL file and the next one because the half-recycled WAL
file was re-recycled upon restarting.  This seems likely to lead to
WAL corruption.

This change replaces all calls to durable_rename_excl() with
durable_rename().  This removes the protection against
accidentally overwriting an existing file, but some platforms are
already living without it, and ordinarily there shouldn't be one.
The function itself is left around in case any extensions are using
it.  It will be removed in v16 via a follow-up commit.

Back-patch to all supported versions.  Before v13,
durable_rename_excl() was named durable_link_or_rename().

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 contrib/basic_archive/basic_archive.c |  5 +++--
 src/backend/access/transam/timeline.c | 14 ++------------
 src/backend/access/transam/xlog.c     |  8 ++------
 3 files changed, 7 insertions(+), 20 deletions(-)

diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index e7efbfb9c3..ed33854c57 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -281,9 +281,10 @@ basic_archive_file_internal(const char *file, const char *path)
 
 	/*
 	 * Sync the temporary file to disk and move it to its final destination.
-	 * This will fail if destination already exists.
+	 * Note that this will overwrite any existing file, but this is only
+	 * possible if someone else created the file since the stat() above.
 	 */
-	(void) durable_rename_excl(temp, destination, ERROR);
+	(void) durable_rename(temp, destination, ERROR);
 
 	ereport(DEBUG1,
 			(errmsg("archived \"%s\" via basic_archive", file)));
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index be21968293..128f754e87 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,7 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -519,12 +514,7 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 61cda56c6f..f49194a8b5 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3323,14 +3323,10 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
-- 
2.25.1

From 398350968f35f0974f1668e06be1adad4a7f7e3c Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 12:38:23 -0700
Subject: [PATCH v3 2/2] Remove durable_rename_excl().

A previous commit replaced all calls to this function with
durable_rename(), but the function itself was not removed in back-
branches since extensions may use it.  This change removes the
function from v16devel.

Do not back-patch.

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 src/backend/storage/file/fd.c  | 63 ----------------------------------
 src/include/pg_config_manual.h |  7 ----
 src/include/storage/fd.h       |  1 -
 3 files changed, 71 deletions(-)

diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 24704b6a02..f904f60c08 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -807,69 +807,6 @@ durable_unlink(const char *fname, int elevel)
 	return 0;
 }
 
-/*
- * durable_rename_excl -- rename a file in a durable manner.
- *
- * Similar to durable_rename(), except that this routine tries (but does not
- * guarantee) not to overwrite the target file.
- *
- * Note that a crash in an unfortunate moment can leave you with two links to
- * the target file.
- *
- * Log errors with the caller specified severity.
- *
- * On Windows, using a hard link followed by unlink() causes concurrency
- * issues, while a simple rename() does not cause that, so be careful when
- * changing the logic of this routine.
- *
- * Returns 0 if the operation succeeded, -1 otherwise. Note that errno is not
- * valid upon return.
- */
-int
-durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
-{
-	/*
-	 * Ensure that, if we crash directly after the rename/link, a file with
-	 * valid contents is moved into place.
-	 */
-	if (fsync_fname_ext(oldfile, false, false, elevel) != 0)
-		return -1;
-
-#ifdef HAVE_WORKING_LINK
-	if (link(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-	unlink(oldfile);
-#else
-	if (rename(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-#endif
-
-	/*
-	 * Make change persistent in case of an OS crash, both the new entry and
-	 * its parent directory need to be flushed.
-	 */
-	if (fsync_fname_ext(newfile, false, false, elevel) != 0)
-		return -1;
-
-	/* Same for parent directory */
-	if (fsync_parent_path(newfile, elevel) != 0)
-		return -1;
-
-	return 0;
-}
-
 /*
  * InitFileAccess --- initialize this module during backend startup
  *
diff --git a/src/include/pg_config_manual.h b/src/include/pg_config_manual.h
index 84ce5a4a5d..830804fdfb 100644
--- a/src/include/pg_config_manual.h
+++ b/src/include/pg_config_manual.h
@@ -163,13 +163,6 @@
 #define USE_BARRIER_SMGRRELEASE
 #endif
 
-/*
- * Define this if your operating system supports link()
- */
-#if !defined(WIN32) && !defined(__CYGWIN__)
-#define HAVE_WORKING_LINK 1
-#endif
-
 /*
  * USE_POSIX_FADVISE controls whether Postgres will attempt to use the
  * posix_fadvise() kernel call.  Usually the automatic configure tests are
diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h
index 69549b000f..2b4a8e0ffe 100644
--- a/src/include/storage/fd.h
+++ b/src/include/storage/fd.h
@@ -187,7 +187,6 @@ extern void fsync_fname(const char *fname, bool isdir);
 extern int	fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel);
 extern int	durable_rename(const char *oldfile, const char *newfile, int loglevel);
 extern int	durable_unlink(const char *fname, int loglevel);
-extern int	durable_rename_excl(const char *oldfile, const char *newfile, int loglevel);
 extern void SyncDataDirectory(void);
 extern int	data_sync_elevel(int elevel);
 
-- 
2.25.1

From 8ffc337621f8a287350a7a55256b58b0585f7a1f Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 11:56:50 -0700
Subject: [PATCH v4 1/2] Replace calls to durable_rename_excl() with
 durable_rename().

durable_rename_excl() attempts to avoid overwriting any existing
files by using link() and unlink(), but it falls back to rename()
on some platforms (e.g., Windows), which offers no such overwrite
protection.  Most callers use durable_rename_excl() just in case
there is an existing file, but in practice there shouldn't be one.
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  My testing demonstrated that
it was possible to end up with two links to the same file in pg_wal
after a crash just before unlink() during WAL recycling.
Specifically, the test produced links to the same file for the
current WAL file and the next one because the half-recycled WAL
file was re-recycled upon restarting.  This seems likely to lead to
WAL corruption.

This change replaces all calls to durable_rename_excl() with
durable_rename().  This removes the protection against
accidentally overwriting an existing file, but some platforms are
already living without it, and ordinarily there shouldn't be one.
The function itself is left around in case any extensions are using
it.  It will be removed in v16 via a follow-up commit.

Back-patch to all supported versions.  Before v13,
durable_rename_excl() was named durable_link_or_rename().

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 contrib/basic_archive/basic_archive.c |  5 +++--
 src/backend/access/transam/timeline.c | 16 ++++------------
 src/backend/access/transam/xlog.c     |  9 +++------
 3 files changed, 10 insertions(+), 20 deletions(-)

diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index e7efbfb9c3..ed33854c57 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -281,9 +281,10 @@ basic_archive_file_internal(const char *file, const char *path)
 
 	/*
 	 * Sync the temporary file to disk and move it to its final destination.
-	 * This will fail if destination already exists.
+	 * Note that this will overwrite any existing file, but this is only
+	 * possible if someone else created the file since the stat() above.
 	 */
-	(void) durable_rename_excl(temp, destination, ERROR);
+	(void) durable_rename(temp, destination, ERROR);
 
 	ereport(DEBUG1,
 			(errmsg("archived \"%s\" via basic_archive", file)));
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index be21968293..f3a8e53aa4 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,8 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -519,12 +515,8 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 61cda56c6f..76ec80c950 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3323,14 +3323,11 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
-- 
2.25.1

From 1d8e3c2040d1acf04e8c14637debf0dc12d500d2 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 12:38:23 -0700
Subject: [PATCH v4 2/2] Remove durable_rename_excl().

A previous commit replaced all calls to this function with
durable_rename(), but the function itself was not removed in back-
branches since extensions may use it.  This change removes the
function from v16devel.

Do not back-patch.

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 src/backend/storage/file/fd.c  | 63 ----------------------------------
 src/include/pg_config_manual.h |  7 ----
 src/include/storage/fd.h       |  1 -
 3 files changed, 71 deletions(-)

diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 24704b6a02..f904f60c08 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -807,69 +807,6 @@ durable_unlink(const char *fname, int elevel)
 	return 0;
 }
 
-/*
- * durable_rename_excl -- rename a file in a durable manner.
- *
- * Similar to durable_rename(), except that this routine tries (but does not
- * guarantee) not to overwrite the target file.
- *
- * Note that a crash in an unfortunate moment can leave you with two links to
- * the target file.
- *
- * Log errors with the caller specified severity.
- *
- * On Windows, using a hard link followed by unlink() causes concurrency
- * issues, while a simple rename() does not cause that, so be careful when
- * changing the logic of this routine.
- *
- * Returns 0 if the operation succeeded, -1 otherwise. Note that errno is not
- * valid upon return.
- */
-int
-durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
-{
-	/*
-	 * Ensure that, if we crash directly after the rename/link, a file with
-	 * valid contents is moved into place.
-	 */
-	if (fsync_fname_ext(oldfile, false, false, elevel) != 0)
-		return -1;
-
-#ifdef HAVE_WORKING_LINK
-	if (link(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-	unlink(oldfile);
-#else
-	if (rename(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-#endif
-
-	/*
-	 * Make change persistent in case of an OS crash, both the new entry and
-	 * its parent directory need to be flushed.
-	 */
-	if (fsync_fname_ext(newfile, false, false, elevel) != 0)
-		return -1;
-
-	/* Same for parent directory */
-	if (fsync_parent_path(newfile, elevel) != 0)
-		return -1;
-
-	return 0;
-}
-
 /*
  * InitFileAccess --- initialize this module during backend startup
  *
diff --git a/src/include/pg_config_manual.h b/src/include/pg_config_manual.h
index 84ce5a4a5d..830804fdfb 100644
--- a/src/include/pg_config_manual.h
+++ b/src/include/pg_config_manual.h
@@ -163,13 +163,6 @@
 #define USE_BARRIER_SMGRRELEASE
 #endif
 
-/*
- * Define this if your operating system supports link()
- */
-#if !defined(WIN32) && !defined(__CYGWIN__)
-#define HAVE_WORKING_LINK 1
-#endif
-
 /*
  * USE_POSIX_FADVISE controls whether Postgres will attempt to use the
  * posix_fadvise() kernel call.  Usually the automatic configure tests are
diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h
index 69549b000f..2b4a8e0ffe 100644
--- a/src/include/storage/fd.h
+++ b/src/include/storage/fd.h
@@ -187,7 +187,6 @@ extern void fsync_fname(const char *fname, bool isdir);
 extern int	fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel);
 extern int	durable_rename(const char *oldfile, const char *newfile, int loglevel);
 extern int	durable_unlink(const char *fname, int loglevel);
-extern int	durable_rename_excl(const char *oldfile, const char *newfile, int loglevel);
 extern void SyncDataDirectory(void);
 extern int	data_sync_elevel(int elevel);
 
-- 
2.25.1

From 312860b4dc3794428c1e31528c05cf4b1ad06f00 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 11:56:50 -0700
Subject: [PATCH v5 1/2] Replace calls to durable_rename_excl() with
 durable_rename().

durable_rename_excl() attempts to avoid overwriting any existing
files by using link() and unlink(), but it falls back to rename()
on some platforms (e.g., Windows), which offers no such overwrite
protection.  Most callers use durable_rename_excl() just in case
there is an existing file, but in practice there shouldn't be one.
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  My testing demonstrated that
it was possible to end up with two links to the same file in pg_wal
after a crash just before unlink() during WAL recycling.
Specifically, the test produced links to the same file for the
current WAL file and the next one because the half-recycled WAL
file was re-recycled upon restarting.  This seems likely to lead to
WAL corruption.

This change replaces all calls to durable_rename_excl() with
durable_rename().  This removes the protection against
accidentally overwriting an existing file, but some platforms are
already living without it, and ordinarily there shouldn't be one.
The function itself is left around in case any extensions are using
it.  It will be removed in v16 via a follow-up commit.

Back-patch to all supported versions.  Before v13,
durable_rename_excl() was named durable_link_or_rename().

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 contrib/basic_archive/basic_archive.c |  5 +++--
 src/backend/access/transam/timeline.c | 18 +++++-------------
 src/backend/access/transam/xlog.c     |  9 +++------
 3 files changed, 11 insertions(+), 21 deletions(-)

diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index e7efbfb9c3..ed33854c57 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -281,9 +281,10 @@ basic_archive_file_internal(const char *file, const char *path)
 
 	/*
 	 * Sync the temporary file to disk and move it to its final destination.
-	 * This will fail if destination already exists.
+	 * Note that this will overwrite any existing file, but this is only
+	 * possible if someone else created the file since the stat() above.
 	 */
-	(void) durable_rename_excl(temp, destination, ERROR);
+	(void) durable_rename(temp, destination, ERROR);
 
 	ereport(DEBUG1,
 			(errmsg("archived \"%s\" via basic_archive", file)));
diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index be21968293..e0a2a8ea68 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,8 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -516,15 +512,11 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the completed history file into place with its final name.
+	 * Now move the completed history file into place with its final name,
+	 * replacing any existing file with the same name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 61cda56c6f..76ec80c950 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3323,14 +3323,11 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
-- 
2.25.1

From b239f1be33ea0e54b8e661312df52af49c090882 Mon Sep 17 00:00:00 2001
From: Nathan Bossart <nathandbossart@gmail.com>
Date: Tue, 26 Apr 2022 12:38:23 -0700
Subject: [PATCH v5 2/2] Remove durable_rename_excl().

A previous commit replaced all calls to this function with
durable_rename(), but the function itself was not removed in back-
branches since extensions may use it.  This change removes the
function from v16devel.

Do not back-patch.

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220418182336.GA2298576%40nathanxps13
---
 src/backend/storage/file/fd.c  | 63 ----------------------------------
 src/include/pg_config_manual.h |  7 ----
 src/include/storage/fd.h       |  1 -
 3 files changed, 71 deletions(-)

diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 24704b6a02..f904f60c08 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -807,69 +807,6 @@ durable_unlink(const char *fname, int elevel)
 	return 0;
 }
 
-/*
- * durable_rename_excl -- rename a file in a durable manner.
- *
- * Similar to durable_rename(), except that this routine tries (but does not
- * guarantee) not to overwrite the target file.
- *
- * Note that a crash in an unfortunate moment can leave you with two links to
- * the target file.
- *
- * Log errors with the caller specified severity.
- *
- * On Windows, using a hard link followed by unlink() causes concurrency
- * issues, while a simple rename() does not cause that, so be careful when
- * changing the logic of this routine.
- *
- * Returns 0 if the operation succeeded, -1 otherwise. Note that errno is not
- * valid upon return.
- */
-int
-durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
-{
-	/*
-	 * Ensure that, if we crash directly after the rename/link, a file with
-	 * valid contents is moved into place.
-	 */
-	if (fsync_fname_ext(oldfile, false, false, elevel) != 0)
-		return -1;
-
-#ifdef HAVE_WORKING_LINK
-	if (link(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-	unlink(oldfile);
-#else
-	if (rename(oldfile, newfile) < 0)
-	{
-		ereport(elevel,
-				(errcode_for_file_access(),
-				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile)));
-		return -1;
-	}
-#endif
-
-	/*
-	 * Make change persistent in case of an OS crash, both the new entry and
-	 * its parent directory need to be flushed.
-	 */
-	if (fsync_fname_ext(newfile, false, false, elevel) != 0)
-		return -1;
-
-	/* Same for parent directory */
-	if (fsync_parent_path(newfile, elevel) != 0)
-		return -1;
-
-	return 0;
-}
-
 /*
  * InitFileAccess --- initialize this module during backend startup
  *
diff --git a/src/include/pg_config_manual.h b/src/include/pg_config_manual.h
index 84ce5a4a5d..830804fdfb 100644
--- a/src/include/pg_config_manual.h
+++ b/src/include/pg_config_manual.h
@@ -163,13 +163,6 @@
 #define USE_BARRIER_SMGRRELEASE
 #endif
 
-/*
- * Define this if your operating system supports link()
- */
-#if !defined(WIN32) && !defined(__CYGWIN__)
-#define HAVE_WORKING_LINK 1
-#endif
-
 /*
  * USE_POSIX_FADVISE controls whether Postgres will attempt to use the
  * posix_fadvise() kernel call.  Usually the automatic configure tests are
diff --git a/src/include/storage/fd.h b/src/include/storage/fd.h
index 69549b000f..2b4a8e0ffe 100644
--- a/src/include/storage/fd.h
+++ b/src/include/storage/fd.h
@@ -187,7 +187,6 @@ extern void fsync_fname(const char *fname, bool isdir);
 extern int	fsync_fname_ext(const char *fname, bool isdir, bool ignore_perm, int elevel);
 extern int	durable_rename(const char *oldfile, const char *newfile, int loglevel);
 extern int	durable_unlink(const char *fname, int loglevel);
-extern int	durable_rename_excl(const char *oldfile, const char *newfile, int loglevel);
 extern void SyncDataDirectory(void);
 extern int	data_sync_elevel(int elevel);
 
-- 
2.25.1

From 168c7071541d231fc13aee424709524f2ed4976e Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Tue, 5 Jul 2022 10:16:12 +0900
Subject: [PATCH] Replace durable_rename_excl() by durable_rename(), take two

durable_rename_excl() attempts to avoid overwriting any existing files
by using link() and unlink(), and it falls back to rename() on some
platforms (aka WIN32), which offers no such overwrite protection.  Most
callers use durable_rename_excl() just in case there is an existing
file, but in practice there shouldn't be one (see below for more
details).

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  As per Nathan's tests, it is
possible to end up with two links to the same file in pg_wal after a
crash just before unlink() during WAL recycling.  Specifically, the test
produced links to the same file for the current WAL file and the next
one because the half-recycled WAL file was re-recycled upon restarting,
leading to WAL corruption.

This change replaces all the calls of durable_rename_excl() to
durable_rename().  This removes the protection against accidentally
overwriting an existing file, but some platforms are already living
without it and ordinarily there shouldn't be one.  The function itself
is left around in case any extensions are using it.  It will be removed
on HEAD via a follow-up commit.

Here is a summary of the existing callers of durable_rename_excl() (see
second discussion link at the bottom), replaced by this commit.  First,
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.  Second, xlog.c uses it to recycle
past WAL segments, where an overwrite should not happen (origin of the
change at f0e37a8) because there are protections about the WAL segment
to select when recycling an entry.  The third and last area is related
to the write of timeline history files.  writeTimeLineHistory() will
write a new timeline history file at the end of recovery on promotion,
so there should be no such files for the same timeline.
What remains is writeTimeLineHistoryFile(), that can be used in parallel
by a WAL receiver and the startup process, and some digging of the
buildfarm shows that EEXIST from a WAL receiver can happen with an error
of "could not link file \"pg_wal/xlogtemp.NN\" to \"pg_wal/MM.history\",
which would cause an automatic restart of the WAL receiver as it is
promoted to FATAL, hence this should improve the stability of the WAL
receiver as rename() would overwrite an existing TLI history file
already fetched by the startup process at recovery.

This is a bug fix, but knowing the unlikeliness of the problem involving
one or more crashes at an exceptionally bad moment, no backpatch is
done.  Also, I want to be careful with such changes (aaa3aed did the
opposite of this change by removing HAVE_WORKING_LINK so as Windows
would do a link() rather than a rename() but this was not
concurrent-safe).  A backpatch could be revisited in the future.  This
is the second time this change is attempted, ccfbd92 being the first
one, but this time no assertions are added for the case of a TLI history
file written concurrently by the WAL receiver or the startup process
because we can expect one to exist (some of the TAP tests are able to
trigger with a proper timing).

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220407182954.GA1231544@nathanxps13
Discussion: https://postgr.es/m/Ym6GZbqQdlalSKSG@paquier.xyz
---
 src/backend/access/transam/timeline.c | 18 +++++-------------
 src/backend/access/transam/xlog.c     |  9 +++------
 contrib/basic_archive/basic_archive.c |  5 +++--
 3 files changed, 11 insertions(+), 21 deletions(-)

diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index be21968293..e0a2a8ea68 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,8 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -516,15 +512,11 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the completed history file into place with its final name.
+	 * Now move the completed history file into place with its final name,
+	 * replacing any existing file with the same name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 5d8322fbd0..818284759d 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3326,14 +3326,11 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
diff --git a/contrib/basic_archive/basic_archive.c b/contrib/basic_archive/basic_archive.c
index 87dd77cdd3..db5fd87429 100644
--- a/contrib/basic_archive/basic_archive.c
+++ b/contrib/basic_archive/basic_archive.c
@@ -282,9 +282,10 @@ basic_archive_file_internal(const char *file, const char *path)
 
 	/*
 	 * Sync the temporary file to disk and move it to its final destination.
-	 * This will fail if destination already exists.
+	 * Note that this will overwrite any existing file, but this is only
+	 * possible if someone else created the file since the stat() above.
 	 */
-	(void) durable_rename_excl(temp, destination, ERROR);
+	(void) durable_rename(temp, destination, ERROR);
 
 	ereport(DEBUG1,
 			(errmsg("archived \"%s\" via basic_archive", file)));
-- 
2.45.2

From 51649664dd7549ea2f7cadc26d024a3ea56d684a Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Tue, 5 Jul 2022 10:16:12 +0900
Subject: [PATCH] Replace durable_rename_excl() by durable_rename(), take two

durable_rename_excl() attempts to avoid overwriting any existing files
by using link() and unlink(), and it falls back to rename() on some
platforms (aka WIN32), which offers no such overwrite protection.  Most
callers use durable_rename_excl() just in case there is an existing
file, but in practice there shouldn't be one (see below for more
details).

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  As per Nathan's tests, it is
possible to end up with two links to the same file in pg_wal after a
crash just before unlink() during WAL recycling.  Specifically, the test
produced links to the same file for the current WAL file and the next
one because the half-recycled WAL file was re-recycled upon restarting,
leading to WAL corruption.

This change replaces all the calls of durable_rename_excl() to
durable_rename().  This removes the protection against accidentally
overwriting an existing file, but some platforms are already living
without it and ordinarily there shouldn't be one.  The function itself
is left around in case any extensions are using it.  It will be removed
on HEAD via a follow-up commit.

Here is a summary of the existing callers of durable_rename_excl() (see
second discussion link at the bottom), replaced by this commit.  First,
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.  Second, xlog.c uses it to recycle
past WAL segments, where an overwrite should not happen (origin of the
change at f0e37a8) because there are protections about the WAL segment
to select when recycling an entry.  The third and last area is related
to the write of timeline history files.  writeTimeLineHistory() will
write a new timeline history file at the end of recovery on promotion,
so there should be no such files for the same timeline.
What remains is writeTimeLineHistoryFile(), that can be used in parallel
by a WAL receiver and the startup process, and some digging of the
buildfarm shows that EEXIST from a WAL receiver can happen with an error
of "could not link file \"pg_wal/xlogtemp.NN\" to \"pg_wal/MM.history\",
which would cause an automatic restart of the WAL receiver as it is
promoted to FATAL, hence this should improve the stability of the WAL
receiver as rename() would overwrite an existing TLI history file
already fetched by the startup process at recovery.

This is a bug fix, but knowing the unlikeliness of the problem involving
one or more crashes at an exceptionally bad moment, no backpatch is
done.  Also, I want to be careful with such changes (aaa3aed did the
opposite of this change by removing HAVE_WORKING_LINK so as Windows
would do a link() rather than a rename() but this was not
concurrent-safe).  A backpatch could be revisited in the future.  This
is the second time this change is attempted, ccfbd92 being the first
one, but this time no assertions are added for the case of a TLI history
file written concurrently by the WAL receiver or the startup process
because we can expect one to exist (some of the TAP tests are able to
trigger with a proper timing).

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220407182954.GA1231544@nathanxps13
Discussion: https://postgr.es/m/Ym6GZbqQdlalSKSG@paquier.xyz
---
 src/backend/access/transam/timeline.c | 18 +++++-------------
 src/backend/access/transam/xlog.c     |  9 +++------
 2 files changed, 8 insertions(+), 19 deletions(-)

diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index 8d0903c175..eeb505fb6a 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,8 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -516,15 +512,11 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the completed history file into place with its final name.
+	 * Now move the completed history file into place with its final name,
+	 * replacing any existing file with the same name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index f1a795bba9..a6e2cb88f3 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3693,15 +3693,12 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		if (use_lock)
 			LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
-- 
2.45.2

From 93c566230e5686da1b1d40186f31b79339be2bcf Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Tue, 5 Jul 2022 10:16:12 +0900
Subject: [PATCH] Replace durable_rename_excl() by durable_rename(), take two

durable_rename_excl() attempts to avoid overwriting any existing files
by using link() and unlink(), and it falls back to rename() on some
platforms (aka WIN32), which offers no such overwrite protection.  Most
callers use durable_rename_excl() just in case there is an existing
file, but in practice there shouldn't be one (see below for more
details).

Furthermore, failures during durable_rename_excl() can result in
multiple hard links to the same file.  As per Nathan's tests, it is
possible to end up with two links to the same file in pg_wal after a
crash just before unlink() during WAL recycling.  Specifically, the test
produced links to the same file for the current WAL file and the next
one because the half-recycled WAL file was re-recycled upon restarting,
leading to WAL corruption.

This change replaces all the calls of durable_rename_excl() to
durable_rename().  This removes the protection against accidentally
overwriting an existing file, but some platforms are already living
without it and ordinarily there shouldn't be one.  The function itself
is left around in case any extensions are using it.  It will be removed
on HEAD via a follow-up commit.

Here is a summary of the existing callers of durable_rename_excl() (see
second discussion link at the bottom), replaced by this commit.  First,
basic_archive used it to avoid overwriting an archive concurrently
created by another server, but as mentioned above, it will still
overwrite files on some platforms.  Second, xlog.c uses it to recycle
past WAL segments, where an overwrite should not happen (origin of the
change at f0e37a8) because there are protections about the WAL segment
to select when recycling an entry.  The third and last area is related
to the write of timeline history files.  writeTimeLineHistory() will
write a new timeline history file at the end of recovery on promotion,
so there should be no such files for the same timeline.
What remains is writeTimeLineHistoryFile(), that can be used in parallel
by a WAL receiver and the startup process, and some digging of the
buildfarm shows that EEXIST from a WAL receiver can happen with an error
of "could not link file \"pg_wal/xlogtemp.NN\" to \"pg_wal/MM.history\",
which would cause an automatic restart of the WAL receiver as it is
promoted to FATAL, hence this should improve the stability of the WAL
receiver as rename() would overwrite an existing TLI history file
already fetched by the startup process at recovery.

This is a bug fix, but knowing the unlikeliness of the problem involving
one or more crashes at an exceptionally bad moment, no backpatch is
done.  Also, I want to be careful with such changes (aaa3aed did the
opposite of this change by removing HAVE_WORKING_LINK so as Windows
would do a link() rather than a rename() but this was not
concurrent-safe).  A backpatch could be revisited in the future.  This
is the second time this change is attempted, ccfbd92 being the first
one, but this time no assertions are added for the case of a TLI history
file written concurrently by the WAL receiver or the startup process
because we can expect one to exist (some of the TAP tests are able to
trigger with a proper timing).

Author: Nathan Bossart
Reviewed-by: Robert Haas, Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/20220407182954.GA1231544@nathanxps13
Discussion: https://postgr.es/m/Ym6GZbqQdlalSKSG@paquier.xyz
---
 src/backend/access/transam/timeline.c | 18 +++++-------------
 src/backend/access/transam/xlog.c     |  9 +++------
 2 files changed, 8 insertions(+), 19 deletions(-)

diff --git a/src/backend/access/transam/timeline.c b/src/backend/access/transam/timeline.c
index e6a29d9a9b..517ab023a3 100644
--- a/src/backend/access/transam/timeline.c
+++ b/src/backend/access/transam/timeline.c
@@ -441,12 +441,8 @@ writeTimeLineHistory(TimeLineID newTLI, TimeLineID parentTLI,
 	 * Now move the completed history file into place with its final name.
 	 */
 	TLHistoryFilePath(path, newTLI);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	durable_rename(tmppath, path, ERROR);
 
 	/* The history file can be archived immediately. */
 	if (XLogArchivingActive())
@@ -516,15 +512,11 @@ writeTimeLineHistoryFile(TimeLineID tli, char *content, int size)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the completed history file into place with its final name.
+	 * Now move the completed history file into place with its final name,
+	 * replacing any existing file with the same name.
 	 */
 	TLHistoryFilePath(path, tli);
-
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	durable_rename_excl(tmppath, path, ERROR);
+	durable_rename(tmppath, path, ERROR);
 }
 
 /*
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index ab4a510ea7..1fa15a58be 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3648,15 +3648,12 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 		}
 	}
 
-	/*
-	 * Perform the rename using link if available, paranoidly trying to avoid
-	 * overwriting an existing file (there shouldn't be one).
-	 */
-	if (durable_rename_excl(tmppath, path, LOG) != 0)
+	Assert(access(path, F_OK) != 0 && errno == ENOENT);
+	if (durable_rename(tmppath, path, LOG) != 0)
 	{
 		if (use_lock)
 			LWLockRelease(ControlFileLock);
-		/* durable_rename_excl already emitted log message */
+		/* durable_rename already emitted log message */
 		return false;
 	}
 
-- 
2.45.2

From 27d2e964c585a2309bd340d856fe9e2402587b0e Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:55 -0700
Subject: [PATCH v14 1/7] Remove XLogFileInit() ability to skip
 ControlFileLock.

Cold paths, initdb and end-of-recovery, used it.  Don't optimize them.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 46 ++++++++-------------------
 src/backend/replication/walreceiver.c |  2 +-
 3 files changed, 16 insertions(+), 34 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index ee3e369b79f3..596965d353d0 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -296,7 +296,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *use_existent, bool use_lock);
+extern int	XLogFileInit(XLogSegNo segno, bool *use_existent);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 0334c2d3bba2..20449820c7cb 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -927,8 +927,7 @@ static void AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic);
 static bool XLogCheckpointNeeded(XLogSegNo new_segno);
 static void XLogWrite(XLogwrtRqst WriteRqst, bool flexible);
 static bool InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
-								   bool find_free, XLogSegNo max_segno,
-								   bool use_lock);
+								   bool find_free, XLogSegNo max_segno);
 static int	XLogFileRead(XLogSegNo segno, int emode, TimeLineID tli,
 						 XLogSource source, bool notfoundOk);
 static int	XLogFileReadAnyTLI(XLogSegNo segno, int emode, XLogSource source);
@@ -2520,7 +2519,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 
 			/* create/use new log file */
 			use_existent = true;
-			openLogFile = XLogFileInit(openLogSegNo, &use_existent, true);
+			openLogFile = XLogFileInit(openLogSegNo, &use_existent);
 			ReserveExternalFD();
 		}
 
@@ -3293,10 +3292,6 @@ XLogNeedsFlush(XLogRecPtr record)
  * pre-existing file will be deleted).  On return, true if a pre-existing
  * file was used.
  *
- * use_lock: if true, acquire ControlFileLock while moving file into
- * place.  This should be true except during bootstrap log creation.  The
- * caller must *not* hold the lock at call.
- *
  * Returns FD of opened file.
  *
  * Note: errors here are ERROR not PANIC because we might or might not be
@@ -3305,7 +3300,7 @@ XLogNeedsFlush(XLogRecPtr record)
  * in a critical section.
  */
 int
-XLogFileInit(XLogSegNo logsegno, bool *use_existent, bool use_lock)
+XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 {
 	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
@@ -3465,8 +3460,7 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent, bool use_lock)
 	 */
 	max_segno = logsegno + CheckPointSegments;
 	if (!InstallXLogFileSegment(&installed_segno, tmppath,
-								*use_existent, max_segno,
-								use_lock))
+								*use_existent, max_segno))
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
@@ -3623,7 +3617,7 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
 	/*
 	 * Now move the segment into place with its final name.
 	 */
-	if (!InstallXLogFileSegment(&destsegno, tmppath, false, 0, false))
+	if (!InstallXLogFileSegment(&destsegno, tmppath, false, 0))
 		elog(ERROR, "InstallXLogFileSegment should not have failed");
 }
 
@@ -3647,29 +3641,20 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
  * free slot is found between *segno and max_segno. (Ignored when find_free
  * is false.)
  *
- * use_lock: if true, acquire ControlFileLock while moving file into
- * place.  This should be true except during bootstrap log creation.  The
- * caller must *not* hold the lock at call.
- *
  * Returns true if the file was installed successfully.  false indicates that
  * max_segno limit was exceeded, or an error occurred while renaming the
  * file into place.
  */
 static bool
 InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
-					   bool find_free, XLogSegNo max_segno,
-					   bool use_lock)
+					   bool find_free, XLogSegNo max_segno)
 {
 	char		path[MAXPGPATH];
 	struct stat stat_buf;
 
 	XLogFilePath(path, ThisTimeLineID, *segno, wal_segment_size);
 
-	/*
-	 * We want to be sure that only one process does this at a time.
-	 */
-	if (use_lock)
-		LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
 
 	if (!find_free)
 	{
@@ -3684,8 +3669,7 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 			if ((*segno) >= max_segno)
 			{
 				/* Failed to find a free slot within specified range */
-				if (use_lock)
-					LWLockRelease(ControlFileLock);
+				LWLockRelease(ControlFileLock);
 				return false;
 			}
 			(*segno)++;
@@ -3696,14 +3680,12 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 	Assert(access(path, F_OK) != 0 && errno == ENOENT);
 	if (durable_rename(tmppath, path, LOG) != 0)
 	{
-		if (use_lock)
-			LWLockRelease(ControlFileLock);
+		LWLockRelease(ControlFileLock);
 		/* durable_rename already emitted log message */
 		return false;
 	}
 
-	if (use_lock)
-		LWLockRelease(ControlFileLock);
+	LWLockRelease(ControlFileLock);
 
 	return true;
 }
@@ -3974,7 +3956,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	{
 		_logSegNo++;
 		use_existent = true;
-		lf = XLogFileInit(_logSegNo, &use_existent, true);
+		lf = XLogFileInit(_logSegNo, &use_existent);
 		close(lf);
 		if (!use_existent)
 			CheckpointStats.ckpt_segs_added++;
@@ -4251,7 +4233,7 @@ RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
 		*endlogSegNo <= recycleSegNo &&
 		lstat(path, &statbuf) == 0 && S_ISREG(statbuf.st_mode) &&
 		InstallXLogFileSegment(endlogSegNo, path,
-							   true, recycleSegNo, true))
+							   true, recycleSegNo))
 	{
 		ereport(DEBUG2,
 				(errmsg_internal("recycled write-ahead log file \"%s\"",
@@ -5388,7 +5370,7 @@ BootStrapXLOG(void)
 
 	/* Create first XLOG segment file */
 	use_existent = false;
-	openLogFile = XLogFileInit(1, &use_existent, false);
+	openLogFile = XLogFileInit(1, &use_existent);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5697,7 +5679,7 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		bool		use_existent = true;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &use_existent, true);
+		fd = XLogFileInit(startLogSegNo, &use_existent);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 4831a259c48a..4ad20f12b61b 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -894,7 +894,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &use_existent, true);
+			recvFile = XLogFileInit(recvSegNo, &use_existent);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From 019002a9e1975c101b7c84f62d89cf48a136c851 Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:55 -0700
Subject: [PATCH v14 2/7] In XLogFileInit(), fix *use_existent postcondition to
 suit callers.

Infrequently, the mismatch caused log_checkpoints messages and
TRACE_POSTGRESQL_CHECKPOINT_DONE() to witness an "added" count too high
by one.  Since that consequence is so minor, no back-patch.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/backend/access/transam/xlog.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 20449820c7cb..c9176eea68c4 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3289,8 +3289,8 @@ XLogNeedsFlush(XLogRecPtr record)
  * logsegno: identify segment to be created/opened.
  *
  * *use_existent: if true, OK to use a pre-existing file (else, any
- * pre-existing file will be deleted).  On return, true if a pre-existing
- * file was used.
+ * pre-existing file will be deleted).  On return, false iff this call added
+ * some segment on disk.
  *
  * Returns FD of opened file.
  *
@@ -3459,8 +3459,10 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	 * CheckPointSegments.
 	 */
 	max_segno = logsegno + CheckPointSegments;
-	if (!InstallXLogFileSegment(&installed_segno, tmppath,
-								*use_existent, max_segno))
+	if (InstallXLogFileSegment(&installed_segno, tmppath,
+							   *use_existent, max_segno))
+		*use_existent = false;
+	else
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
@@ -3470,9 +3472,6 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 		unlink(tmppath);
 	}
 
-	/* Set flag to tell caller there was no existent file */
-	*use_existent = false;
-
 	/* Now open original target segment (might not be file I just made) */
 	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
 	if (fd < 0)
-- 
2.47.2

From 8f4517030288192c81e32ddf456028dade87104e Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v14 3/7] Remove XLogFileInit() ability to unlink a
 pre-existing file.

Only initdb used it.  initdb refuses to operate on a non-empty directory
and generally does not cope with pre-existing files of other kinds.
Hence, use the opportunity to simplify.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 61 +++++++++++----------------
 src/backend/replication/walreceiver.c |  4 +-
 3 files changed, 28 insertions(+), 39 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 596965d353d0..1860107f9a2e 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -296,7 +296,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *use_existent);
+extern int	XLogFileInit(XLogSegNo segno, bool *added);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index c9176eea68c4..d86295470908 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -2452,7 +2452,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	bool		ispartialpage;
 	bool		last_iteration;
 	bool		finishing_seg;
-	bool		use_existent;
+	bool		added;
 	int			curridx;
 	int			npages;
 	int			startidx;
@@ -2518,8 +2518,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 							wal_segment_size);
 
 			/* create/use new log file */
-			use_existent = true;
-			openLogFile = XLogFileInit(openLogSegNo, &use_existent);
+			openLogFile = XLogFileInit(openLogSegNo, &added);
 			ReserveExternalFD();
 		}
 
@@ -3288,9 +3287,7 @@ XLogNeedsFlush(XLogRecPtr record)
  *
  * logsegno: identify segment to be created/opened.
  *
- * *use_existent: if true, OK to use a pre-existing file (else, any
- * pre-existing file will be deleted).  On return, false iff this call added
- * some segment on disk.
+ * *added: on return, true if this call raised the number of extant segments.
  *
  * Returns FD of opened file.
  *
@@ -3300,7 +3297,7 @@ XLogNeedsFlush(XLogRecPtr record)
  * in a critical section.
  */
 int
-XLogFileInit(XLogSegNo logsegno, bool *use_existent)
+XLogFileInit(XLogSegNo logsegno, bool *added)
 {
 	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
@@ -3315,19 +3312,17 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	/*
 	 * Try to use existent file (checkpoint maker may have created it already)
 	 */
-	if (*use_existent)
+	*added = false;
+	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
+	if (fd < 0)
 	{
-		fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
-		if (fd < 0)
-		{
-			if (errno != ENOENT)
-				ereport(ERROR,
-						(errcode_for_file_access(),
-						 errmsg("could not open file \"%s\": %m", path)));
-		}
-		else
-			return fd;
+		if (errno != ENOENT)
+			ereport(ERROR,
+					(errcode_for_file_access(),
+					 errmsg("could not open file \"%s\": %m", path)));
 	}
+	else
+		return fd;
 
 	/*
 	 * Initialize an empty (all zeroes) segment.  NOTE: it is possible that
@@ -3440,12 +3435,9 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the segment into place with its final name.
-	 *
-	 * If caller didn't want to use a pre-existing file, get rid of any
-	 * pre-existing file.  Otherwise, cope with possibility that someone else
-	 * has created the file while we were filling ours: if so, use ours to
-	 * pre-create a future log segment.
+	 * Now move the segment into place with its final name.  Cope with
+	 * possibility that someone else has created the file while we were
+	 * filling ours: if so, use ours to pre-create a future log segment.
 	 */
 	installed_segno = logsegno;
 
@@ -3459,9 +3451,8 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	 * CheckPointSegments.
 	 */
 	max_segno = logsegno + CheckPointSegments;
-	if (InstallXLogFileSegment(&installed_segno, tmppath,
-							   *use_existent, max_segno))
-		*use_existent = false;
+	if (InstallXLogFileSegment(&installed_segno, tmppath, true, max_segno))
+		*added = true;
 	else
 	{
 		/*
@@ -3946,7 +3937,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 {
 	XLogSegNo	_logSegNo;
 	int			lf;
-	bool		use_existent;
+	bool		added;
 	uint64		offset;
 
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
@@ -3954,10 +3945,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	if (offset >= (uint32) (0.75 * wal_segment_size))
 	{
 		_logSegNo++;
-		use_existent = true;
-		lf = XLogFileInit(_logSegNo, &use_existent);
+		lf = XLogFileInit(_logSegNo, &added);
 		close(lf);
-		if (!use_existent)
+		if (added)
 			CheckpointStats.ckpt_segs_added++;
 	}
 }
@@ -5271,7 +5261,7 @@ BootStrapXLOG(void)
 	XLogLongPageHeader longpage;
 	XLogRecord *record;
 	char	   *recptr;
-	bool		use_existent;
+	bool		added;
 	uint64		sysidentifier;
 	struct timeval tv;
 	pg_crc32c	crc;
@@ -5368,8 +5358,7 @@ BootStrapXLOG(void)
 	record->xl_crc = crc;
 
 	/* Create first XLOG segment file */
-	use_existent = false;
-	openLogFile = XLogFileInit(1, &use_existent);
+	openLogFile = XLogFileInit(1, &added);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5675,10 +5664,10 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		 * The switch happened at a segment boundary, so just create the next
 		 * segment on the new timeline.
 		 */
-		bool		use_existent = true;
+		bool		added;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &use_existent);
+		fd = XLogFileInit(startLogSegNo, &added);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 4ad20f12b61b..b6f07ab81f1d 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -890,11 +890,11 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 		if (recvFile < 0)
 		{
-			bool		use_existent = true;
+			bool		added = true;
 
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &use_existent);
+			recvFile = XLogFileInit(recvSegNo, &added);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From 3086175a74b701e1ae1ddf6423cf796b6e2bde4c Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Mon, 10 Mar 2025 13:18:52 +0900
Subject: [PATCH v14 4/7] Revert "Add HINT for restartpoint race with
 KeepFileRestoredFromArchive()."

This reverts commit 8ad6c5dbbe5a234c55c6663020db297251756006.
---
 src/backend/access/transam/xlog.c |  5 +----
 src/backend/storage/file/fd.c     | 10 ++--------
 2 files changed, 3 insertions(+), 12 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index d86295470908..793451bdf70c 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3468,10 +3468,7 @@ XLogFileInit(XLogSegNo logsegno, bool *added)
 	if (fd < 0)
 		ereport(ERROR,
 				(errcode_for_file_access(),
-				 errmsg("could not open file \"%s\": %m", path),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+				 errmsg("could not open file \"%s\": %m", path)));
 
 	elog(DEBUG2, "done creating and filling new WAL file");
 
diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 09c4a93e40b0..484a07302a40 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -834,10 +834,7 @@ durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
 		ereport(elevel,
 				(errcode_for_file_access(),
 				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+						oldfile, newfile)));
 		return -1;
 	}
 	unlink(oldfile);
@@ -847,10 +844,7 @@ durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
 		ereport(elevel,
 				(errcode_for_file_access(),
 				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+						oldfile, newfile)));
 		return -1;
 	}
 #endif
-- 
2.47.2

From 712671efd7b77fada6dad308dbaaba28cab22064 Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v14 5/7] Don't ERROR on PreallocXlogFiles() race condition.

Before a restartpoint finishes PreallocXlogFiles(), a startup process
KeepFileRestoredFromArchive() call can unlink the preallocated segment.
If a CHECKPOINT sql command had elicited the restartpoint experiencing
the race condition, that sql command failed.  Moreover, the restartpoint
omitted its log_checkpoints message and some inessential resource
reclamation.  Prevent the ERROR by skipping open() of the segment.
Since these consequences are so minor, no back-patch.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 79 +++++++++++++++++++--------
 src/backend/replication/walreceiver.c |  4 +-
 3 files changed, 58 insertions(+), 27 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 1860107f9a2e..a0706c06a791 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -296,7 +296,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *added);
+extern int	XLogFileInit(XLogSegNo segno);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 793451bdf70c..73a828f075ad 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -2452,7 +2452,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	bool		ispartialpage;
 	bool		last_iteration;
 	bool		finishing_seg;
-	bool		added;
 	int			curridx;
 	int			npages;
 	int			startidx;
@@ -2518,7 +2517,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 							wal_segment_size);
 
 			/* create/use new log file */
-			openLogFile = XLogFileInit(openLogSegNo, &added);
+			openLogFile = XLogFileInit(openLogSegNo);
 			ReserveExternalFD();
 		}
 
@@ -3283,23 +3282,21 @@ XLogNeedsFlush(XLogRecPtr record)
 }
 
 /*
- * Create a new XLOG file segment, or open a pre-existing one.
+ * Try to make a given XLOG file segment exist.
  *
- * logsegno: identify segment to be created/opened.
+ * logsegno: identify segment.
  *
  * *added: on return, true if this call raised the number of extant segments.
  *
- * Returns FD of opened file.
+ * path: on return, this char[MAXPGPATH] has the path to the logsegno file.
  *
- * Note: errors here are ERROR not PANIC because we might or might not be
- * inside a critical section (eg, during checkpoint there is no reason to
- * take down the system on failure).  They will promote to PANIC if we are
- * in a critical section.
+ * Returns -1 or FD of opened file.  A -1 here is not an error; a caller
+ * wanting an open segment should attempt to open "path", which usually will
+ * succeed.  (This is weird, but it's efficient for the callers.)
  */
-int
-XLogFileInit(XLogSegNo logsegno, bool *added)
+static int
+XLogFileInitInternal(XLogSegNo logsegno, bool *added, char *path)
 {
-	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
 	PGAlignedXLogBlock zbuffer;
 	XLogSegNo	installed_segno;
@@ -3452,26 +3449,53 @@ XLogFileInit(XLogSegNo logsegno, bool *added)
 	 */
 	max_segno = logsegno + CheckPointSegments;
 	if (InstallXLogFileSegment(&installed_segno, tmppath, true, max_segno))
+	{
 		*added = true;
+		elog(DEBUG2, "done creating and filling new WAL file");
+	}
 	else
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
-		 * failed to rename the file into place. If the rename failed, opening
-		 * the file below will fail.
+		 * failed to rename the file into place. If the rename failed, a
+		 * caller opening the file may fail.
 		 */
 		unlink(tmppath);
+		elog(DEBUG2, "abandoned new WAL file");
 	}
 
+	return -1;
+}
+
+/*
+ * Create a new XLOG file segment, or open a pre-existing one.
+ *
+ * logsegno: identify segment to be created/opened.
+ *
+ * Returns FD of opened file.
+ *
+ * Note: errors here are ERROR not PANIC because we might or might not be
+ * inside a critical section (eg, during checkpoint there is no reason to
+ * take down the system on failure).  They will promote to PANIC if we are
+ * in a critical section.
+ */
+int
+XLogFileInit(XLogSegNo logsegno)
+{
+	bool		ignore_added;
+	char		path[MAXPGPATH];
+	int			fd;
+
+	fd = XLogFileInitInternal(logsegno, &ignore_added, path);
+	if (fd >= 0)
+		return fd;
+
 	/* Now open original target segment (might not be file I just made) */
 	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
 	if (fd < 0)
 		ereport(ERROR,
 				(errcode_for_file_access(),
 				 errmsg("could not open file \"%s\": %m", path)));
-
-	elog(DEBUG2, "done creating and filling new WAL file");
-
 	return fd;
 }
 
@@ -3928,6 +3952,15 @@ XLogFileClose(void)
  * High-volume systems will be OK once they've built up a sufficient set of
  * recycled log segments, but the startup transient is likely to include
  * a lot of segment creations by foreground processes, which is not so good.
+ *
+ * XLogFileInitInternal() can ereport(ERROR).  All known causes indicate big
+ * trouble; for example, a full filesystem is one cause.  The checkpoint WAL
+ * and/or ControlFile updates already completed.  If a RequestCheckpoint()
+ * initiated the present checkpoint and an ERROR ends this function, the
+ * command that called RequestCheckpoint() fails.  That's not ideal, but it's
+ * not worth contorting more functions to use caller-specified elevel values.
+ * (With or without RequestCheckpoint(), an ERROR forestalls some inessential
+ * reporting and resource reclamation.)
  */
 static void
 PreallocXlogFiles(XLogRecPtr endptr)
@@ -3935,6 +3968,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	XLogSegNo	_logSegNo;
 	int			lf;
 	bool		added;
+	char		path[MAXPGPATH];
 	uint64		offset;
 
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
@@ -3942,8 +3976,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	if (offset >= (uint32) (0.75 * wal_segment_size))
 	{
 		_logSegNo++;
-		lf = XLogFileInit(_logSegNo, &added);
-		close(lf);
+		lf = XLogFileInitInternal(_logSegNo, &added, path);
+		if (lf >= 0)
+			close(lf);
 		if (added)
 			CheckpointStats.ckpt_segs_added++;
 	}
@@ -5258,7 +5293,6 @@ BootStrapXLOG(void)
 	XLogLongPageHeader longpage;
 	XLogRecord *record;
 	char	   *recptr;
-	bool		added;
 	uint64		sysidentifier;
 	struct timeval tv;
 	pg_crc32c	crc;
@@ -5355,7 +5389,7 @@ BootStrapXLOG(void)
 	record->xl_crc = crc;
 
 	/* Create first XLOG segment file */
-	openLogFile = XLogFileInit(1, &added);
+	openLogFile = XLogFileInit(1);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5661,10 +5695,9 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		 * The switch happened at a segment boundary, so just create the next
 		 * segment on the new timeline.
 		 */
-		bool		added;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &added);
+		fd = XLogFileInit(startLogSegNo);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index b6f07ab81f1d..7ae04b9682c7 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -890,11 +890,9 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 		if (recvFile < 0)
 		{
-			bool		added = true;
-
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &added);
+			recvFile = XLogFileInit(recvSegNo);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From 8bf97022fdd53a019c709efb757a5f1d953a1474 Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v14 6/7] Skip WAL recycling and preallocation during archive
 recovery.

The previous commit addressed the chief consequences of a race condition
between InstallXLogFileSegment() and KeepFileRestoredFromArchive().  Fix
three lesser consequences.  A spurious durable_rename_excl() LOG message
remained possible.  KeepFileRestoredFromArchive() wasted the proceeds of
WAL recycling and preallocation.  Finally, XLogFileInitInternal() could
return a descriptor for a file that KeepFileRestoredFromArchive() had
already unlinked.  That felt like a recipe for future bugs.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/backend/access/transam/xlog.c | 65 +++++++++++++++++++++++++++----
 1 file changed, 57 insertions(+), 8 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 73a828f075ad..c5fe7e6fa4cc 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -673,6 +673,16 @@ typedef struct XLogCtlData
 	 */
 	bool		SharedHotStandbyActive;
 
+	/*
+	 * InstallXLogFileSegmentActive indicates whether the checkpointer should
+	 * arrange for future segments by recycling and/or PreallocXlogFiles().
+	 * Protected by ControlFileLock.  Only the startup process changes it.  If
+	 * true, anyone can use InstallXLogFileSegment().  If false, the startup
+	 * process owns the exclusive right to install segments, by reading from
+	 * the archive and possibly replacing existing files.
+	 */
+	bool		InstallXLogFileSegmentActive;
+
 	/*
 	 * SharedPromoteIsTriggered indicates if a standby promotion has been
 	 * triggered.  Protected by info_lck.
@@ -935,6 +945,7 @@ static int	XLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr,
 						 int reqLen, XLogRecPtr targetRecPtr, char *readBuf);
 static bool WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 										bool fetching_ckpt, XLogRecPtr tliRecPtr);
+static void XLogShutdownWalRcv(void);
 static int	emode_for_corrupt_record(int emode, XLogRecPtr RecPtr);
 static void XLogFileClose(void);
 static void PreallocXlogFiles(XLogRecPtr endptr);
@@ -3653,8 +3664,8 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
  * is false.)
  *
  * Returns true if the file was installed successfully.  false indicates that
- * max_segno limit was exceeded, or an error occurred while renaming the
- * file into place.
+ * max_segno limit was exceeded, the startup process has disabled this
+ * function for now, or an error occurred while renaming the file into place.
  */
 static bool
 InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
@@ -3666,6 +3677,11 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 	XLogFilePath(path, ThisTimeLineID, *segno, wal_segment_size);
 
 	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	if (!XLogCtl->InstallXLogFileSegmentActive)
+	{
+		LWLockRelease(ControlFileLock);
+		return false;
+	}
 
 	if (!find_free)
 	{
@@ -3770,6 +3786,7 @@ XLogFileRead(XLogSegNo segno, int emode, TimeLineID tli,
 	 */
 	if (source == XLOG_FROM_ARCHIVE)
 	{
+		Assert(!XLogCtl->InstallXLogFileSegmentActive);
 		KeepFileRestoredFromArchive(path, xlogfname);
 
 		/*
@@ -3971,6 +3988,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	char		path[MAXPGPATH];
 	uint64		offset;
 
+	if (!XLogCtl->InstallXLogFileSegmentActive)
+		return;					/* unlocked check says no */
+
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
 	offset = XLogSegmentOffset(endptr - 1, wal_segment_size);
 	if (offset >= (uint32) (0.75 * wal_segment_size))
@@ -4252,6 +4272,7 @@ RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
 	 */
 	if (wal_recycle &&
 		*endlogSegNo <= recycleSegNo &&
+		XLogCtl->InstallXLogFileSegmentActive &&	/* callee rechecks this */
 		lstat(path, &statbuf) == 0 && S_ISREG(statbuf.st_mode) &&
 		InstallXLogFileSegment(endlogSegNo, path,
 							   true, recycleSegNo))
@@ -4265,7 +4286,7 @@ RemoveXlogFile(const char *segname, XLogSegNo recycleSegNo,
 	}
 	else
 	{
-		/* No need for any more future segments... */
+		/* No need for any more future segments, or recycling failed ... */
 		int			rc;
 
 		ereport(DEBUG2,
@@ -5270,6 +5291,7 @@ XLOGShmemInit(void)
 	XLogCtl->XLogCacheBlck = XLOGbuffers - 1;
 	XLogCtl->SharedRecoveryState = RECOVERY_STATE_CRASH;
 	XLogCtl->SharedHotStandbyActive = false;
+	XLogCtl->InstallXLogFileSegmentActive = false;
 	XLogCtl->SharedPromoteIsTriggered = false;
 	XLogCtl->WalWriterSleeping = false;
 
@@ -5297,6 +5319,11 @@ BootStrapXLOG(void)
 	struct timeval tv;
 	pg_crc32c	crc;
 
+	/* allow ordinary WAL segment creation, like StartupXLOG() would */
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = true;
+	LWLockRelease(ControlFileLock);
+
 	/*
 	 * Select a hopefully-unique system identifier code for this installation.
 	 * We use the result of gettimeofday(), including the fractional seconds
@@ -7685,7 +7712,7 @@ StartupXLOG(void)
 	 * over these records and subsequent ones if it's still alive when we
 	 * start writing WAL.
 	 */
-	ShutdownWalRcv();
+	XLogShutdownWalRcv();
 
 	/*
 	 * Reset unlogged relations to the contents of their INIT fork. This is
@@ -7710,7 +7737,7 @@ StartupXLOG(void)
 	 * recovery, e.g., timeline history file) from archive or pg_wal.
 	 *
 	 * Note that standby mode must be turned off after killing WAL receiver,
-	 * i.e., calling ShutdownWalRcv().
+	 * i.e., calling XLogShutdownWalRcv().
 	 */
 	Assert(!WalRcvStreaming());
 	StandbyMode = false;
@@ -7779,6 +7806,14 @@ StartupXLOG(void)
 	 */
 	oldestActiveXID = PrescanPreparedTransactions(NULL, NULL);
 
+	/*
+	 * Allow ordinary WAL segment creation before any exitArchiveRecovery(),
+	 * which sometimes creates a segment, and after the last ReadRecord().
+	 */
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = true;
+	LWLockRelease(ControlFileLock);
+
 	/*
 	 * Consider whether we need to assign a new timeline ID.
 	 *
@@ -12717,7 +12752,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 */
 					if (StandbyMode && CheckForStandbyTrigger())
 					{
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 						return false;
 					}
 
@@ -12765,7 +12800,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 * WAL that we restore from archive.
 					 */
 					if (WalRcvStreaming())
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 
 					/*
 					 * Before we sleep, re-scan for possible new timelines if
@@ -12895,7 +12930,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 */
 					if (pendingWalRcvRestart && !startWalReceiver)
 					{
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 
 						/*
 						 * Re-scan for possible new timelines if we were
@@ -12945,6 +12980,9 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 									 tli, curFileTLI);
 						}
 						curFileTLI = tli;
+						LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+						XLogCtl->InstallXLogFileSegmentActive = true;
+						LWLockRelease(ControlFileLock);
 						RequestXLogStreaming(tli, ptr, PrimaryConnInfo,
 											 PrimarySlotName,
 											 wal_receiver_create_temp_slot);
@@ -13115,6 +13153,17 @@ StartupRequestWalReceiverRestart(void)
 	}
 }
 
+/* Thin wrapper around ShutdownWalRcv(). */
+static void
+XLogShutdownWalRcv(void)
+{
+	ShutdownWalRcv();
+
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = false;
+	LWLockRelease(ControlFileLock);
+}
+
 /*
  * Determine what log level should be used to report a corrupt WAL record
  * in the current WAL page, previously read by XLogPageRead().
-- 
2.47.2

From c2d49a28a5d1a0e48c7a9edf97c481d984af98e9 Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Thu, 15 Sep 2022 06:45:23 -0700
Subject: [PATCH v14 7/7] Reset InstallXLogFileSegmentActive after walreceiver
 self-initiated exit.

After commit cc2c7d65fc27e877c9f407587b0b92d46cd6dd16 added this flag,
failure to reset it caused assertion failures.  In non-assert builds, it
made the system fail to achieve the objectives listed in that commit;
chiefly, we might emit a spurious log message.  Back-patch to v15, where
that commit first appeared.

Bharath Rupireddy and Kyotaro Horiguchi.  Reviewed by Dilip Kumar,
Nathan Bossart and Michael Paquier.  Reported by Dilip Kumar.

Discussion: https://postgr.es/m/CAFiTN-sE3ry=ycMPVtC+Djw4Fd7gbUGVv_qqw6qfzp=JLvqT3g@mail.gmail.com
---
 src/backend/access/transam/xlog.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index c5fe7e6fa4cc..9ce4297e3de6 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -12799,8 +12799,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 * walreceiver is not active, so that it won't overwrite
 					 * WAL that we restore from archive.
 					 */
-					if (WalRcvStreaming())
-						XLogShutdownWalRcv();
+					XLogShutdownWalRcv();
 
 					/*
 					 * Before we sleep, re-scan for possible new timelines if
-- 
2.47.2

From d5de431953286182b8b0bfea4c0ecf23c1d31fbc Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:55 -0700
Subject: [PATCH v13 1/7] Remove XLogFileInit() ability to skip
 ControlFileLock.

Cold paths, initdb and end-of-recovery, used it.  Don't optimize them.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 46 ++++++++-------------------
 src/backend/replication/walreceiver.c |  2 +-
 3 files changed, 16 insertions(+), 34 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 01b3b0a767dc..7cb147a7ee37 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -286,7 +286,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *use_existent, bool use_lock);
+extern int	XLogFileInit(XLogSegNo segno, bool *use_existent);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 769f4a202f28..859f86cfc3cd 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -917,8 +917,7 @@ static void AdvanceXLInsertBuffer(XLogRecPtr upto, bool opportunistic);
 static bool XLogCheckpointNeeded(XLogSegNo new_segno);
 static void XLogWrite(XLogwrtRqst WriteRqst, bool flexible);
 static bool InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
-								   bool find_free, XLogSegNo max_segno,
-								   bool use_lock);
+								   bool find_free, XLogSegNo max_segno);
 static int	XLogFileRead(XLogSegNo segno, int emode, TimeLineID tli,
 						 XLogSource source, bool notfoundOk);
 static int	XLogFileReadAnyTLI(XLogSegNo segno, int emode, XLogSource source);
@@ -2509,7 +2508,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 
 			/* create/use new log file */
 			use_existent = true;
-			openLogFile = XLogFileInit(openLogSegNo, &use_existent, true);
+			openLogFile = XLogFileInit(openLogSegNo, &use_existent);
 			ReserveExternalFD();
 		}
 
@@ -3263,10 +3262,6 @@ XLogNeedsFlush(XLogRecPtr record)
  * pre-existing file will be deleted).  On return, true if a pre-existing
  * file was used.
  *
- * use_lock: if true, acquire ControlFileLock while moving file into
- * place.  This should be true except during bootstrap log creation.  The
- * caller must *not* hold the lock at call.
- *
  * Returns FD of opened file.
  *
  * Note: errors here are ERROR not PANIC because we might or might not be
@@ -3275,7 +3270,7 @@ XLogNeedsFlush(XLogRecPtr record)
  * in a critical section.
  */
 int
-XLogFileInit(XLogSegNo logsegno, bool *use_existent, bool use_lock)
+XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 {
 	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
@@ -3420,8 +3415,7 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent, bool use_lock)
 	 */
 	max_segno = logsegno + CheckPointSegments;
 	if (!InstallXLogFileSegment(&installed_segno, tmppath,
-								*use_existent, max_segno,
-								use_lock))
+								*use_existent, max_segno))
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
@@ -3578,7 +3572,7 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
 	/*
 	 * Now move the segment into place with its final name.
 	 */
-	if (!InstallXLogFileSegment(&destsegno, tmppath, false, 0, false))
+	if (!InstallXLogFileSegment(&destsegno, tmppath, false, 0))
 		elog(ERROR, "InstallXLogFileSegment should not have failed");
 }
 
@@ -3602,29 +3596,20 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
  * free slot is found between *segno and max_segno. (Ignored when find_free
  * is false.)
  *
- * use_lock: if true, acquire ControlFileLock while moving file into
- * place.  This should be true except during bootstrap log creation.  The
- * caller must *not* hold the lock at call.
- *
  * Returns true if the file was installed successfully.  false indicates that
  * max_segno limit was exceeded, or an error occurred while renaming the
  * file into place.
  */
 static bool
 InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
-					   bool find_free, XLogSegNo max_segno,
-					   bool use_lock)
+					   bool find_free, XLogSegNo max_segno)
 {
 	char		path[MAXPGPATH];
 	struct stat stat_buf;
 
 	XLogFilePath(path, ThisTimeLineID, *segno, wal_segment_size);
 
-	/*
-	 * We want to be sure that only one process does this at a time.
-	 */
-	if (use_lock)
-		LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
 
 	if (!find_free)
 	{
@@ -3639,8 +3624,7 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 			if ((*segno) >= max_segno)
 			{
 				/* Failed to find a free slot within specified range */
-				if (use_lock)
-					LWLockRelease(ControlFileLock);
+				LWLockRelease(ControlFileLock);
 				return false;
 			}
 			(*segno)++;
@@ -3651,14 +3635,12 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 	Assert(access(path, F_OK) != 0 && errno == ENOENT);
 	if (durable_rename(tmppath, path, LOG) != 0)
 	{
-		if (use_lock)
-			LWLockRelease(ControlFileLock);
+		LWLockRelease(ControlFileLock);
 		/* durable_rename already emitted log message */
 		return false;
 	}
 
-	if (use_lock)
-		LWLockRelease(ControlFileLock);
+	LWLockRelease(ControlFileLock);
 
 	return true;
 }
@@ -3929,7 +3911,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	{
 		_logSegNo++;
 		use_existent = true;
-		lf = XLogFileInit(_logSegNo, &use_existent, true);
+		lf = XLogFileInit(_logSegNo, &use_existent);
 		close(lf);
 		if (!use_existent)
 			CheckpointStats.ckpt_segs_added++;
@@ -4206,7 +4188,7 @@ RemoveXlogFile(const char *segname, XLogRecPtr lastredoptr, XLogRecPtr endptr)
 		endlogSegNo <= recycleSegNo &&
 		lstat(path, &statbuf) == 0 && S_ISREG(statbuf.st_mode) &&
 		InstallXLogFileSegment(&endlogSegNo, path,
-							   true, recycleSegNo, true))
+							   true, recycleSegNo))
 	{
 		ereport(DEBUG2,
 				(errmsg("recycled write-ahead log file \"%s\"",
@@ -5342,7 +5324,7 @@ BootStrapXLOG(void)
 
 	/* Create first XLOG segment file */
 	use_existent = false;
-	openLogFile = XLogFileInit(1, &use_existent, false);
+	openLogFile = XLogFileInit(1, &use_existent);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5651,7 +5633,7 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		bool		use_existent = true;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &use_existent, true);
+		fd = XLogFileInit(startLogSegNo, &use_existent);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 479c6c441bda..b874cf364f06 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -922,7 +922,7 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &use_existent, true);
+			recvFile = XLogFileInit(recvSegNo, &use_existent);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From e805d4e8b2bcfb7d72552ce914e480606c978bef Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:55 -0700
Subject: [PATCH v13 2/7] In XLogFileInit(), fix *use_existent postcondition to
 suit callers.

Infrequently, the mismatch caused log_checkpoints messages and
TRACE_POSTGRESQL_CHECKPOINT_DONE() to witness an "added" count too high
by one.  Since that consequence is so minor, no back-patch.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/backend/access/transam/xlog.c | 13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 859f86cfc3cd..28fd6da7d366 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3259,8 +3259,8 @@ XLogNeedsFlush(XLogRecPtr record)
  * logsegno: identify segment to be created/opened.
  *
  * *use_existent: if true, OK to use a pre-existing file (else, any
- * pre-existing file will be deleted).  On return, true if a pre-existing
- * file was used.
+ * pre-existing file will be deleted).  On return, false iff this call added
+ * some segment on disk.
  *
  * Returns FD of opened file.
  *
@@ -3414,8 +3414,10 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	 * CheckPointSegments.
 	 */
 	max_segno = logsegno + CheckPointSegments;
-	if (!InstallXLogFileSegment(&installed_segno, tmppath,
-								*use_existent, max_segno))
+	if (InstallXLogFileSegment(&installed_segno, tmppath,
+							   *use_existent, max_segno))
+		*use_existent = false;
+	else
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
@@ -3425,9 +3427,6 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 		unlink(tmppath);
 	}
 
-	/* Set flag to tell caller there was no existent file */
-	*use_existent = false;
-
 	/* Now open original target segment (might not be file I just made) */
 	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
 	if (fd < 0)
-- 
2.47.2

From 3944585703be84b60e7ff1c45c2578ee312aa736 Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v13 3/7] Remove XLogFileInit() ability to unlink a
 pre-existing file.

Only initdb used it.  initdb refuses to operate on a non-empty directory
and generally does not cope with pre-existing files of other kinds.
Hence, use the opportunity to simplify.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 61 +++++++++++----------------
 src/backend/replication/walreceiver.c |  4 +-
 3 files changed, 28 insertions(+), 39 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 7cb147a7ee37..6119452ba8cd 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -286,7 +286,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *use_existent);
+extern int	XLogFileInit(XLogSegNo segno, bool *added);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 28fd6da7d366..6fbb5328fd0b 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -2440,7 +2440,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	bool		ispartialpage;
 	bool		last_iteration;
 	bool		finishing_seg;
-	bool		use_existent;
+	bool		added;
 	int			curridx;
 	int			npages;
 	int			startidx;
@@ -2507,8 +2507,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 							wal_segment_size);
 
 			/* create/use new log file */
-			use_existent = true;
-			openLogFile = XLogFileInit(openLogSegNo, &use_existent);
+			openLogFile = XLogFileInit(openLogSegNo, &added);
 			ReserveExternalFD();
 		}
 
@@ -3258,9 +3257,7 @@ XLogNeedsFlush(XLogRecPtr record)
  *
  * logsegno: identify segment to be created/opened.
  *
- * *use_existent: if true, OK to use a pre-existing file (else, any
- * pre-existing file will be deleted).  On return, false iff this call added
- * some segment on disk.
+ * *added: on return, true if this call raised the number of extant segments.
  *
  * Returns FD of opened file.
  *
@@ -3270,7 +3267,7 @@ XLogNeedsFlush(XLogRecPtr record)
  * in a critical section.
  */
 int
-XLogFileInit(XLogSegNo logsegno, bool *use_existent)
+XLogFileInit(XLogSegNo logsegno, bool *added)
 {
 	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
@@ -3286,19 +3283,17 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	/*
 	 * Try to use existent file (checkpoint maker may have created it already)
 	 */
-	if (*use_existent)
+	*added = false;
+	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
+	if (fd < 0)
 	{
-		fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
-		if (fd < 0)
-		{
-			if (errno != ENOENT)
-				ereport(ERROR,
-						(errcode_for_file_access(),
-						 errmsg("could not open file \"%s\": %m", path)));
-		}
-		else
-			return fd;
+		if (errno != ENOENT)
+			ereport(ERROR,
+					(errcode_for_file_access(),
+					 errmsg("could not open file \"%s\": %m", path)));
 	}
+	else
+		return fd;
 
 	/*
 	 * Initialize an empty (all zeroes) segment.  NOTE: it is possible that
@@ -3395,12 +3390,9 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 				 errmsg("could not close file \"%s\": %m", tmppath)));
 
 	/*
-	 * Now move the segment into place with its final name.
-	 *
-	 * If caller didn't want to use a pre-existing file, get rid of any
-	 * pre-existing file.  Otherwise, cope with possibility that someone else
-	 * has created the file while we were filling ours: if so, use ours to
-	 * pre-create a future log segment.
+	 * Now move the segment into place with its final name.  Cope with
+	 * possibility that someone else has created the file while we were
+	 * filling ours: if so, use ours to pre-create a future log segment.
 	 */
 	installed_segno = logsegno;
 
@@ -3414,9 +3406,8 @@ XLogFileInit(XLogSegNo logsegno, bool *use_existent)
 	 * CheckPointSegments.
 	 */
 	max_segno = logsegno + CheckPointSegments;
-	if (InstallXLogFileSegment(&installed_segno, tmppath,
-							   *use_existent, max_segno))
-		*use_existent = false;
+	if (InstallXLogFileSegment(&installed_segno, tmppath, true, max_segno))
+		*added = true;
 	else
 	{
 		/*
@@ -3901,7 +3892,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 {
 	XLogSegNo	_logSegNo;
 	int			lf;
-	bool		use_existent;
+	bool		added;
 	uint64		offset;
 
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
@@ -3909,10 +3900,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	if (offset >= (uint32) (0.75 * wal_segment_size))
 	{
 		_logSegNo++;
-		use_existent = true;
-		lf = XLogFileInit(_logSegNo, &use_existent);
+		lf = XLogFileInit(_logSegNo, &added);
 		close(lf);
-		if (!use_existent)
+		if (added)
 			CheckpointStats.ckpt_segs_added++;
 	}
 }
@@ -5225,7 +5215,7 @@ BootStrapXLOG(void)
 	XLogLongPageHeader longpage;
 	XLogRecord *record;
 	char	   *recptr;
-	bool		use_existent;
+	bool		added;
 	uint64		sysidentifier;
 	struct timeval tv;
 	pg_crc32c	crc;
@@ -5322,8 +5312,7 @@ BootStrapXLOG(void)
 	record->xl_crc = crc;
 
 	/* Create first XLOG segment file */
-	use_existent = false;
-	openLogFile = XLogFileInit(1, &use_existent);
+	openLogFile = XLogFileInit(1, &added);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5629,10 +5618,10 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		 * The switch happened at a segment boundary, so just create the next
 		 * segment on the new timeline.
 		 */
-		bool		use_existent = true;
+		bool		added;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &use_existent);
+		fd = XLogFileInit(startLogSegNo, &added);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index b874cf364f06..976dbc04cde7 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -918,11 +918,11 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 		if (recvFile < 0)
 		{
-			bool		use_existent = true;
+			bool		added = true;
 
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &use_existent);
+			recvFile = XLogFileInit(recvSegNo, &added);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From 994fd5e49f76e40b9eae6d02d0a463d5048e9302 Mon Sep 17 00:00:00 2001
From: Michael Paquier <michael@paquier.xyz>
Date: Mon, 10 Mar 2025 13:18:52 +0900
Subject: [PATCH v13 4/7] Revert "Add HINT for restartpoint race with
 KeepFileRestoredFromArchive()."

This reverts commit 8ad6c5dbbe5a234c55c6663020db297251756006.
---
 src/backend/access/transam/xlog.c |  5 +----
 src/backend/storage/file/fd.c     | 10 ++--------
 2 files changed, 3 insertions(+), 12 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 6fbb5328fd0b..4ef417f1eddb 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -3423,10 +3423,7 @@ XLogFileInit(XLogSegNo logsegno, bool *added)
 	if (fd < 0)
 		ereport(ERROR,
 				(errcode_for_file_access(),
-				 errmsg("could not open file \"%s\": %m", path),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+				 errmsg("could not open file \"%s\": %m", path)));
 
 	elog(DEBUG2, "done creating and filling new WAL file");
 
diff --git a/src/backend/storage/file/fd.c b/src/backend/storage/file/fd.c
index 452ab88cafad..7aa50a5d2e57 100644
--- a/src/backend/storage/file/fd.c
+++ b/src/backend/storage/file/fd.c
@@ -800,10 +800,7 @@ durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
 		ereport(elevel,
 				(errcode_for_file_access(),
 				 errmsg("could not link file \"%s\" to \"%s\": %m",
-						oldfile, newfile),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+						oldfile, newfile)));
 		return -1;
 	}
 	unlink(oldfile);
@@ -813,10 +810,7 @@ durable_rename_excl(const char *oldfile, const char *newfile, int elevel)
 		ereport(elevel,
 				(errcode_for_file_access(),
 				 errmsg("could not rename file \"%s\" to \"%s\": %m",
-						oldfile, newfile),
-				 (AmCheckpointerProcess() ?
-				  errhint("This is known to fail occasionally during archive recovery, where it is harmless.") :
-				  0)));
+						oldfile, newfile)));
 		return -1;
 	}
 #endif
-- 
2.47.2

From 1ab26db206a00ce2a4b24996b7b13ae841662dfc Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v13 5/7] Don't ERROR on PreallocXlogFiles() race condition.

Before a restartpoint finishes PreallocXlogFiles(), a startup process
KeepFileRestoredFromArchive() call can unlink the preallocated segment.
If a CHECKPOINT sql command had elicited the restartpoint experiencing
the race condition, that sql command failed.  Moreover, the restartpoint
omitted its log_checkpoints message and some inessential resource
reclamation.  Prevent the ERROR by skipping open() of the segment.
Since these consequences are so minor, no back-patch.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/include/access/xlog.h             |  2 +-
 src/backend/access/transam/xlog.c     | 79 +++++++++++++++++++--------
 src/backend/replication/walreceiver.c |  4 +-
 3 files changed, 58 insertions(+), 27 deletions(-)

diff --git a/src/include/access/xlog.h b/src/include/access/xlog.h
index 6119452ba8cd..7526378d57f0 100644
--- a/src/include/access/xlog.h
+++ b/src/include/access/xlog.h
@@ -286,7 +286,7 @@ extern XLogRecPtr XLogInsertRecord(struct XLogRecData *rdata,
 extern void XLogFlush(XLogRecPtr RecPtr);
 extern bool XLogBackgroundFlush(void);
 extern bool XLogNeedsFlush(XLogRecPtr RecPtr);
-extern int	XLogFileInit(XLogSegNo segno, bool *added);
+extern int	XLogFileInit(XLogSegNo segno);
 extern int	XLogFileOpen(XLogSegNo segno);
 
 extern void CheckXLogRemoved(XLogSegNo segno, TimeLineID tli);
diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 4ef417f1eddb..bbc9dd87b40a 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -2440,7 +2440,6 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 	bool		ispartialpage;
 	bool		last_iteration;
 	bool		finishing_seg;
-	bool		added;
 	int			curridx;
 	int			npages;
 	int			startidx;
@@ -2507,7 +2506,7 @@ XLogWrite(XLogwrtRqst WriteRqst, bool flexible)
 							wal_segment_size);
 
 			/* create/use new log file */
-			openLogFile = XLogFileInit(openLogSegNo, &added);
+			openLogFile = XLogFileInit(openLogSegNo);
 			ReserveExternalFD();
 		}
 
@@ -3253,23 +3252,21 @@ XLogNeedsFlush(XLogRecPtr record)
 }
 
 /*
- * Create a new XLOG file segment, or open a pre-existing one.
+ * Try to make a given XLOG file segment exist.
  *
- * logsegno: identify segment to be created/opened.
+ * logsegno: identify segment.
  *
  * *added: on return, true if this call raised the number of extant segments.
  *
- * Returns FD of opened file.
+ * path: on return, this char[MAXPGPATH] has the path to the logsegno file.
  *
- * Note: errors here are ERROR not PANIC because we might or might not be
- * inside a critical section (eg, during checkpoint there is no reason to
- * take down the system on failure).  They will promote to PANIC if we are
- * in a critical section.
+ * Returns -1 or FD of opened file.  A -1 here is not an error; a caller
+ * wanting an open segment should attempt to open "path", which usually will
+ * succeed.  (This is weird, but it's efficient for the callers.)
  */
-int
-XLogFileInit(XLogSegNo logsegno, bool *added)
+static int
+XLogFileInitInternal(XLogSegNo logsegno, bool *added, char *path)
 {
-	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
 	PGAlignedXLogBlock zbuffer;
 	XLogSegNo	installed_segno;
@@ -3407,26 +3404,53 @@ XLogFileInit(XLogSegNo logsegno, bool *added)
 	 */
 	max_segno = logsegno + CheckPointSegments;
 	if (InstallXLogFileSegment(&installed_segno, tmppath, true, max_segno))
+	{
 		*added = true;
+		elog(DEBUG2, "done creating and filling new WAL file");
+	}
 	else
 	{
 		/*
 		 * No need for any more future segments, or InstallXLogFileSegment()
-		 * failed to rename the file into place. If the rename failed, opening
-		 * the file below will fail.
+		 * failed to rename the file into place. If the rename failed, a
+		 * caller opening the file may fail.
 		 */
 		unlink(tmppath);
+		elog(DEBUG2, "abandoned new WAL file");
 	}
 
+	return -1;
+}
+
+/*
+ * Create a new XLOG file segment, or open a pre-existing one.
+ *
+ * logsegno: identify segment to be created/opened.
+ *
+ * Returns FD of opened file.
+ *
+ * Note: errors here are ERROR not PANIC because we might or might not be
+ * inside a critical section (eg, during checkpoint there is no reason to
+ * take down the system on failure).  They will promote to PANIC if we are
+ * in a critical section.
+ */
+int
+XLogFileInit(XLogSegNo logsegno)
+{
+	bool		ignore_added;
+	char		path[MAXPGPATH];
+	int			fd;
+
+	fd = XLogFileInitInternal(logsegno, &ignore_added, path);
+	if (fd >= 0)
+		return fd;
+
 	/* Now open original target segment (might not be file I just made) */
 	fd = BasicOpenFile(path, O_RDWR | PG_BINARY | get_sync_bit(sync_method));
 	if (fd < 0)
 		ereport(ERROR,
 				(errcode_for_file_access(),
 				 errmsg("could not open file \"%s\": %m", path)));
-
-	elog(DEBUG2, "done creating and filling new WAL file");
-
 	return fd;
 }
 
@@ -3883,6 +3907,15 @@ XLogFileClose(void)
  * High-volume systems will be OK once they've built up a sufficient set of
  * recycled log segments, but the startup transient is likely to include
  * a lot of segment creations by foreground processes, which is not so good.
+ *
+ * XLogFileInitInternal() can ereport(ERROR).  All known causes indicate big
+ * trouble; for example, a full filesystem is one cause.  The checkpoint WAL
+ * and/or ControlFile updates already completed.  If a RequestCheckpoint()
+ * initiated the present checkpoint and an ERROR ends this function, the
+ * command that called RequestCheckpoint() fails.  That's not ideal, but it's
+ * not worth contorting more functions to use caller-specified elevel values.
+ * (With or without RequestCheckpoint(), an ERROR forestalls some inessential
+ * reporting and resource reclamation.)
  */
 static void
 PreallocXlogFiles(XLogRecPtr endptr)
@@ -3890,6 +3923,7 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	XLogSegNo	_logSegNo;
 	int			lf;
 	bool		added;
+	char		path[MAXPGPATH];
 	uint64		offset;
 
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
@@ -3897,8 +3931,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	if (offset >= (uint32) (0.75 * wal_segment_size))
 	{
 		_logSegNo++;
-		lf = XLogFileInit(_logSegNo, &added);
-		close(lf);
+		lf = XLogFileInitInternal(_logSegNo, &added, path);
+		if (lf >= 0)
+			close(lf);
 		if (added)
 			CheckpointStats.ckpt_segs_added++;
 	}
@@ -5212,7 +5247,6 @@ BootStrapXLOG(void)
 	XLogLongPageHeader longpage;
 	XLogRecord *record;
 	char	   *recptr;
-	bool		added;
 	uint64		sysidentifier;
 	struct timeval tv;
 	pg_crc32c	crc;
@@ -5309,7 +5343,7 @@ BootStrapXLOG(void)
 	record->xl_crc = crc;
 
 	/* Create first XLOG segment file */
-	openLogFile = XLogFileInit(1, &added);
+	openLogFile = XLogFileInit(1);
 
 	/*
 	 * We needn't bother with Reserve/ReleaseExternalFD here, since we'll
@@ -5615,10 +5649,9 @@ exitArchiveRecovery(TimeLineID endTLI, XLogRecPtr endOfLog)
 		 * The switch happened at a segment boundary, so just create the next
 		 * segment on the new timeline.
 		 */
-		bool		added;
 		int			fd;
 
-		fd = XLogFileInit(startLogSegNo, &added);
+		fd = XLogFileInit(startLogSegNo);
 
 		if (close(fd) != 0)
 		{
diff --git a/src/backend/replication/walreceiver.c b/src/backend/replication/walreceiver.c
index 976dbc04cde7..e1d7c0f75435 100644
--- a/src/backend/replication/walreceiver.c
+++ b/src/backend/replication/walreceiver.c
@@ -918,11 +918,9 @@ XLogWalRcvWrite(char *buf, Size nbytes, XLogRecPtr recptr)
 
 		if (recvFile < 0)
 		{
-			bool		added = true;
-
 			/* Create/use new log file */
 			XLByteToSeg(recptr, recvSegNo, wal_segment_size);
-			recvFile = XLogFileInit(recvSegNo, &added);
+			recvFile = XLogFileInit(recvSegNo);
 			recvFileTLI = ThisTimeLineID;
 		}
 
-- 
2.47.2

From 74f9824b0a2db22f0fabaa6059d6a8d33dd4e72b Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Mon, 28 Jun 2021 18:34:56 -0700
Subject: [PATCH v13 6/7] Skip WAL recycling and preallocation during archive
 recovery.

The previous commit addressed the chief consequences of a race condition
between InstallXLogFileSegment() and KeepFileRestoredFromArchive().  Fix
three lesser consequences.  A spurious durable_rename_excl() LOG message
remained possible.  KeepFileRestoredFromArchive() wasted the proceeds of
WAL recycling and preallocation.  Finally, XLogFileInitInternal() could
return a descriptor for a file that KeepFileRestoredFromArchive() had
already unlinked.  That felt like a recipe for future bugs.

Discussion: https://postgr.es/m/20210202151416.GB3304930@rfd.leadboat.com
---
 src/backend/access/transam/xlog.c | 65 +++++++++++++++++++++++++++----
 1 file changed, 57 insertions(+), 8 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index bbc9dd87b40a..d585af79a6ec 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -675,6 +675,16 @@ typedef struct XLogCtlData
 	 */
 	bool		SharedHotStandbyActive;
 
+	/*
+	 * InstallXLogFileSegmentActive indicates whether the checkpointer should
+	 * arrange for future segments by recycling and/or PreallocXlogFiles().
+	 * Protected by ControlFileLock.  Only the startup process changes it.  If
+	 * true, anyone can use InstallXLogFileSegment().  If false, the startup
+	 * process owns the exclusive right to install segments, by reading from
+	 * the archive and possibly replacing existing files.
+	 */
+	bool		InstallXLogFileSegmentActive;
+
 	/*
 	 * SharedPromoteIsTriggered indicates if a standby promotion has been
 	 * triggered.  Protected by info_lck.
@@ -925,6 +935,7 @@ static int	XLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr,
 						 int reqLen, XLogRecPtr targetRecPtr, char *readBuf);
 static bool WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 										bool fetching_ckpt, XLogRecPtr tliRecPtr);
+static void XLogShutdownWalRcv(void);
 static int	emode_for_corrupt_record(int emode, XLogRecPtr RecPtr);
 static void XLogFileClose(void);
 static void PreallocXlogFiles(XLogRecPtr endptr);
@@ -3608,8 +3619,8 @@ XLogFileCopy(XLogSegNo destsegno, TimeLineID srcTLI, XLogSegNo srcsegno,
  * is false.)
  *
  * Returns true if the file was installed successfully.  false indicates that
- * max_segno limit was exceeded, or an error occurred while renaming the
- * file into place.
+ * max_segno limit was exceeded, the startup process has disabled this
+ * function for now, or an error occurred while renaming the file into place.
  */
 static bool
 InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
@@ -3621,6 +3632,11 @@ InstallXLogFileSegment(XLogSegNo *segno, char *tmppath,
 	XLogFilePath(path, ThisTimeLineID, *segno, wal_segment_size);
 
 	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	if (!XLogCtl->InstallXLogFileSegmentActive)
+	{
+		LWLockRelease(ControlFileLock);
+		return false;
+	}
 
 	if (!find_free)
 	{
@@ -3725,6 +3741,7 @@ XLogFileRead(XLogSegNo segno, int emode, TimeLineID tli,
 	 */
 	if (source == XLOG_FROM_ARCHIVE)
 	{
+		Assert(!XLogCtl->InstallXLogFileSegmentActive);
 		KeepFileRestoredFromArchive(path, xlogfname);
 
 		/*
@@ -3926,6 +3943,9 @@ PreallocXlogFiles(XLogRecPtr endptr)
 	char		path[MAXPGPATH];
 	uint64		offset;
 
+	if (!XLogCtl->InstallXLogFileSegmentActive)
+		return;					/* unlocked check says no */
+
 	XLByteToPrevSeg(endptr, _logSegNo, wal_segment_size);
 	offset = XLogSegmentOffset(endptr - 1, wal_segment_size);
 	if (offset >= (uint32) (0.75 * wal_segment_size))
@@ -4207,6 +4227,7 @@ RemoveXlogFile(const char *segname, XLogRecPtr lastredoptr, XLogRecPtr endptr)
 	 */
 	if (wal_recycle &&
 		endlogSegNo <= recycleSegNo &&
+		XLogCtl->InstallXLogFileSegmentActive &&	/* callee rechecks this */
 		lstat(path, &statbuf) == 0 && S_ISREG(statbuf.st_mode) &&
 		InstallXLogFileSegment(&endlogSegNo, path,
 							   true, recycleSegNo))
@@ -4220,7 +4241,7 @@ RemoveXlogFile(const char *segname, XLogRecPtr lastredoptr, XLogRecPtr endptr)
 	}
 	else
 	{
-		/* No need for any more future segments... */
+		/* No need for any more future segments, or recycling failed ... */
 		int			rc;
 
 		ereport(DEBUG2,
@@ -5225,6 +5246,7 @@ XLOGShmemInit(void)
 	XLogCtl->XLogCacheBlck = XLOGbuffers - 1;
 	XLogCtl->SharedRecoveryState = RECOVERY_STATE_CRASH;
 	XLogCtl->SharedHotStandbyActive = false;
+	XLogCtl->InstallXLogFileSegmentActive = false;
 	XLogCtl->SharedPromoteIsTriggered = false;
 	XLogCtl->WalWriterSleeping = false;
 
@@ -5251,6 +5273,11 @@ BootStrapXLOG(void)
 	struct timeval tv;
 	pg_crc32c	crc;
 
+	/* allow ordinary WAL segment creation, like StartupXLOG() would */
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = true;
+	LWLockRelease(ControlFileLock);
+
 	/*
 	 * Select a hopefully-unique system identifier code for this installation.
 	 * We use the result of gettimeofday(), including the fractional seconds
@@ -7531,7 +7558,7 @@ StartupXLOG(void)
 	 * over these records and subsequent ones if it's still alive when we
 	 * start writing WAL.
 	 */
-	ShutdownWalRcv();
+	XLogShutdownWalRcv();
 
 	/*
 	 * Reset unlogged relations to the contents of their INIT fork. This is
@@ -7556,7 +7583,7 @@ StartupXLOG(void)
 	 * recovery, e.g., timeline history file) from archive or pg_wal.
 	 *
 	 * Note that standby mode must be turned off after killing WAL receiver,
-	 * i.e., calling ShutdownWalRcv().
+	 * i.e., calling XLogShutdownWalRcv().
 	 */
 	Assert(!WalRcvStreaming());
 	StandbyMode = false;
@@ -7625,6 +7652,14 @@ StartupXLOG(void)
 	 */
 	oldestActiveXID = PrescanPreparedTransactions(NULL, NULL);
 
+	/*
+	 * Allow ordinary WAL segment creation before any exitArchiveRecovery(),
+	 * which sometimes creates a segment, and after the last ReadRecord().
+	 */
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = true;
+	LWLockRelease(ControlFileLock);
+
 	/*
 	 * Consider whether we need to assign a new timeline ID.
 	 *
@@ -12491,7 +12526,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 */
 					if (StandbyMode && CheckForStandbyTrigger())
 					{
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 						return false;
 					}
 
@@ -12539,7 +12574,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 * WAL that we restore from archive.
 					 */
 					if (WalRcvStreaming())
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 
 					/*
 					 * Before we sleep, re-scan for possible new timelines if
@@ -12669,7 +12704,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 */
 					if (pendingWalRcvRestart && !startWalReceiver)
 					{
-						ShutdownWalRcv();
+						XLogShutdownWalRcv();
 
 						/*
 						 * Re-scan for possible new timelines if we were
@@ -12720,6 +12755,9 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 									 tli, curFileTLI);
 						}
 						curFileTLI = tli;
+						LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+						XLogCtl->InstallXLogFileSegmentActive = true;
+						LWLockRelease(ControlFileLock);
 						RequestXLogStreaming(tli, ptr, PrimaryConnInfo,
 											 PrimarySlotName,
 											 wal_receiver_create_temp_slot);
@@ -12882,6 +12920,17 @@ StartupRequestWalReceiverRestart(void)
 	}
 }
 
+/* Thin wrapper around ShutdownWalRcv(). */
+static void
+XLogShutdownWalRcv(void)
+{
+	ShutdownWalRcv();
+
+	LWLockAcquire(ControlFileLock, LW_EXCLUSIVE);
+	XLogCtl->InstallXLogFileSegmentActive = false;
+	LWLockRelease(ControlFileLock);
+}
+
 /*
  * Determine what log level should be used to report a corrupt WAL record
  * in the current WAL page, previously read by XLogPageRead().
-- 
2.47.2

From e1eedb586974de10baf149c128987ad2dde447bc Mon Sep 17 00:00:00 2001
From: Noah Misch <noah@leadboat.com>
Date: Thu, 15 Sep 2022 06:45:23 -0700
Subject: [PATCH v13 7/7] Reset InstallXLogFileSegmentActive after walreceiver
 self-initiated exit.

After commit cc2c7d65fc27e877c9f407587b0b92d46cd6dd16 added this flag,
failure to reset it caused assertion failures.  In non-assert builds, it
made the system fail to achieve the objectives listed in that commit;
chiefly, we might emit a spurious log message.  Back-patch to v15, where
that commit first appeared.

Bharath Rupireddy and Kyotaro Horiguchi.  Reviewed by Dilip Kumar,
Nathan Bossart and Michael Paquier.  Reported by Dilip Kumar.

Discussion: https://postgr.es/m/CAFiTN-sE3ry=ycMPVtC+Djw4Fd7gbUGVv_qqw6qfzp=JLvqT3g@mail.gmail.com
---
 src/backend/access/transam/xlog.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index d585af79a6ec..fc104a377ad9 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -12573,8 +12573,7 @@ WaitForWALToBecomeAvailable(XLogRecPtr RecPtr, bool randAccess,
 					 * walreceiver is not active, so that it won't overwrite
 					 * WAL that we restore from archive.
 					 */
-					if (WalRcvStreaming())
-						XLogShutdownWalRcv();
+					XLogShutdownWalRcv();
 
 					/*
 					 * Before we sleep, re-scan for possible new timelines if
-- 
2.47.2

Author:     Noah Misch <noah@leadboat.com>
Commit:     Noah Misch <noah@leadboat.com>

    Test restartpoints in archive recovery.
    
    v14 commit 1f95181b44c843729caaa688f74babe9403b5850 and its v13 equivalent
    caused timing-dependent failures in archive recovery, at restartpoints.  The
    symptom was "invalid magic number 0000 in log segment X, offset 0",
    "unexpected pageaddr X in log segment Y, offset 0" [X < Y], or an assertion
    failure.  Commit FIXME and predecessors back-patched v15 changes to fix that.
    This test reproduces the problem probabilistically, typically in less than
    1000 iterations of the test.  Hence, buildfarm and CI runs would have surfaced
    enough failures to get attention within a day.  Back-patch to v13 (all
    supported versions).
    
    Reported by Arun Thirupathi.  Reviewed by FIXME.
    
    Discussion: https://postgr.es/m/20250306193013.36.nmisch@google.com

diff --git a/src/test/recovery/meson.build b/src/test/recovery/meson.build
index 057bcde..cb98376 100644
--- a/src/test/recovery/meson.build
+++ b/src/test/recovery/meson.build
@@ -53,6 +53,7 @@ tests += {
       't/042_low_level_backup.pl',
       't/043_no_contrecord_switch.pl',
       't/044_invalidate_inactive_slots.pl',
+      't/045_archive_restartpoint.pl',
     ],
   },
 }
diff --git a/src/test/recovery/t/045_archive_restartpoint.pl b/src/test/recovery/t/045_archive_restartpoint.pl
new file mode 100644
index 0000000..b143bc4
--- /dev/null
+++ b/src/test/recovery/t/045_archive_restartpoint.pl
@@ -0,0 +1,57 @@
+
+# Copyright (c) 2024-2025, PostgreSQL Global Development Group
+
+# Test restartpoints during archive recovery.
+use strict;
+use warnings;
+
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+
+my $archive_max_mb = 320;
+my $wal_segsize = 1;
+
+# Initialize primary node
+my $node_primary = PostgreSQL::Test::Cluster->new('primary');
+$node_primary->init(
+	has_archiving => 1,
+	allows_streaming => 1,
+	extra => [ '--wal-segsize' => $wal_segsize ]);
+$node_primary->start;
+my $backup_name = 'my_backup';
+$node_primary->backup($backup_name);
+
+$node_primary->safe_psql('postgres',
+	('DO $$BEGIN FOR i IN 1..' . $archive_max_mb / $wal_segsize)
+	  . ' LOOP CHECKPOINT; PERFORM pg_switch_wal(); END LOOP; END$$;');
+
+# Force archiving of WAL file containing recovery target
+my $until_lsn = $node_primary->lsn('write');
+$node_primary->safe_psql('postgres', "SELECT pg_switch_wal()");
+$node_primary->stop;
+
+# Archive recovery
+my $node_restore = PostgreSQL::Test::Cluster->new('restore');
+$node_restore->init_from_backup($node_primary, $backup_name,
+	has_restoring => 1);
+$node_restore->append_conf('postgresql.conf',
+	"recovery_target_lsn = '$until_lsn'");
+$node_restore->append_conf('postgresql.conf',
+	'recovery_target_action = pause');
+$node_restore->append_conf('postgresql.conf',
+	'max_wal_size = ' . 2 * $wal_segsize);
+$node_restore->append_conf('postgresql.conf', 'log_checkpoints = on');
+
+$node_restore->start;
+
+# Wait until restore has replayed enough data
+my $caughtup_query =
+  "SELECT '$until_lsn'::pg_lsn <= pg_last_wal_replay_lsn()";
+$node_restore->poll_query_until('postgres', $caughtup_query)
+  or die "Timed out while waiting for restore to catch up";
+
+$node_restore->stop;
+ok(1, 'restore caught up');
+
+done_testing();