Allow I/O reliability checks using 16-bit checksums

Checksums are set immediately prior to flush out of shared buffers
and checked when pages are read in again. Hint bit setting will
require full page write when block is dirtied, which causes various
infrastructure changes. Extensive comments, docs and README.

WARNING message thrown if checksum fails on non-all zeroes page;
ERROR thrown but can be disabled with ignore_checksum_failure = on.

Feature enabled by an initdb option, since transition from option off
to option on is long and complex and has not yet been implemented.
Default is not to use checksums.

Checksum used is WAL CRC-32 truncated to 16-bits.

Simon Riggs, Jeff Davis, Greg Smith
Wide input and assistance from many community members. Thank you.

2 files changed, 32 insertions, 1 deletions

diff --git a/contrib/pg_upgrade/controldata.c b/contrib/pg_upgrade/controldata.c
index c33c20c7c01..41a8c694e7a 100644
--- a/contrib/pg_upgrade/controldata.c
+++ b/contrib/pg_upgrade/controldata.c
@@ -56,6 +56,7 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 	bool		got_toast = false;
 	bool		got_date_is_int = false;
 	bool		got_float8_pass_by_value = false;
+	bool		got_data_checksums = false;
 	char	   *lc_collate = NULL;
 	char	   *lc_ctype = NULL;
 	char	   *lc_monetary = NULL;
@@ -131,6 +132,13 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 		got_float8_pass_by_value = true;
 	}
 
+	/* Only in <= 9.2 */
+	if (GET_MAJOR_VERSION(cluster->major_version) <= 902)
+	{
+		cluster->controldata.data_checksums = false;
+		got_data_checksums = true;
+	}
+
 	/* we have the result of cmd in "output". so parse it line by line now */
 	while (fgets(bufin, sizeof(bufin), output))
 	{
@@ -393,6 +401,18 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 			cluster->controldata.float8_pass_by_value = strstr(p, "by value") != NULL;
 			got_float8_pass_by_value = true;
 		}
+		else if ((p = strstr(bufin, "checksums")) != NULL)
+		{
+			p = strchr(p, ':');
+
+			if (p == NULL || strlen(p) <= 1)
+				pg_log(PG_FATAL, "%d: controldata retrieval problem\n", __LINE__);
+
+			p++;				/* removing ':' char */
+			/* used later for contrib check */
+			cluster->controldata.data_checksums = strstr(p, "enabled") != NULL;
+			got_data_checksums = true;
+		}
 		/* In pre-8.4 only */
 		else if ((p = strstr(bufin, "LC_COLLATE:")) != NULL)
 		{
@@ -476,7 +496,7 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 		!got_tli ||
 		!got_align || !got_blocksz || !got_largesz || !got_walsz ||
 		!got_walseg || !got_ident || !got_index || !got_toast ||
-		!got_date_is_int || !got_float8_pass_by_value)
+		!got_date_is_int || !got_float8_pass_by_value || !got_data_checksums)
 	{
 		pg_log(PG_REPORT,
 			"The %s cluster lacks some required control information:\n",
@@ -535,6 +555,10 @@ get_control_data(ClusterInfo *cluster, bool live_check)
 		if (!got_float8_pass_by_value)
 			pg_log(PG_REPORT, "  float8 argument passing method\n");
 
+		/* value added in Postgres 9.3 */
+		if (!got_data_checksums)
+			pg_log(PG_REPORT, "  data checksums\n");
+
 		pg_log(PG_FATAL,
 			   "Cannot continue without required control information, terminating\n");
 	}
@@ -596,6 +620,12 @@ check_control_data(ControlData *oldctrl,
 			   "--disable-integer-datetimes or get server binaries built with those\n"
 			   "options.\n");
 	}
+
+	if (oldctrl->data_checksums != newctrl->data_checksums)
+	{
+		pg_log(PG_FATAL,
+			   "old and new pg_controldata checksums settings are invalid or do not match\n");
+	}
 }
 
 
diff --git a/contrib/pg_upgrade/pg_upgrade.h b/contrib/pg_upgrade/pg_upgrade.h
index 497098199f6..370315f0cb3 100644
--- a/contrib/pg_upgrade/pg_upgrade.h
+++ b/contrib/pg_upgrade/pg_upgrade.h
@@ -202,6 +202,7 @@ typedef struct
 	uint32		toast;
 	bool		date_is_int;
 	bool		float8_pass_by_value;
+	bool		data_checksums;
 	char	   *lc_collate;
 	char	   *lc_ctype;
 	char	   *encoding;