From 5f93c37805e7485488480916b4585e098d3cc883 Mon Sep 17 00:00:00 2001
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Thu, 5 Jun 2014 11:31:06 -0400
Subject: [PATCH] Add defenses against running with a wrong selection of
 LOBLKSIZE.

It's critical that the backend's idea of LOBLKSIZE match the way data has
actually been divided up in pg_largeobject.  While we don't provide any
direct way to adjust that value, doing so is a one-line source code change
and various people have expressed interest recently in changing it.  So,
just as with TOAST_MAX_CHUNK_SIZE, it seems prudent to record the value in
pg_control and cross-check that the backend's compiled-in setting matches
the on-disk data.

Also tweak the code in inv_api.c so that fetches from pg_largeobject
explicitly verify that the length of the data field is not more than
LOBLKSIZE.  Formerly we just had Asserts() for that, which is no protection
at all in production builds.  In some of the call sites an overlength data
value would translate directly to a security-relevant stack clobber, so it
seems worth one extra runtime comparison to be sure.

In the back branches, we can't change the contents of pg_control; but we
can still make the extra checks in inv_api.c, which will offer some amount
of protection against running with the wrong value of LOBLKSIZE.
---
 src/backend/access/transam/xlog.c          |  9 +++
 src/backend/storage/large_object/inv_api.c | 84 ++++++++++------------
 src/bin/pg_controldata/pg_controldata.c    |  2 +
 src/bin/pg_resetxlog/pg_resetxlog.c        |  4 ++
 src/include/catalog/pg_control.h           |  3 +-
 src/include/storage/large_object.h         |  2 +
 6 files changed, 58 insertions(+), 46 deletions(-)

diff --git a/src/backend/access/transam/xlog.c b/src/backend/access/transam/xlog.c
index 926bb895fa4..9eca63cbbfa 100644
--- a/src/backend/access/transam/xlog.c
+++ b/src/backend/access/transam/xlog.c
@@ -49,6 +49,7 @@
 #include "storage/bufmgr.h"
 #include "storage/fd.h"
 #include "storage/ipc.h"
+#include "storage/large_object.h"
 #include "storage/latch.h"
 #include "storage/pmsignal.h"
 #include "storage/predicate.h"
@@ -4352,6 +4353,7 @@ WriteControlFile(void)
 	ControlFile->indexMaxKeys = INDEX_MAX_KEYS;
 
 	ControlFile->toast_max_chunk_size = TOAST_MAX_CHUNK_SIZE;
+	ControlFile->loblksize = LOBLKSIZE;
 
 #ifdef HAVE_INT64_TIMESTAMP
 	ControlFile->enableIntTimes = true;
@@ -4545,6 +4547,13 @@ ReadControlFile(void)
 				" but the server was compiled with TOAST_MAX_CHUNK_SIZE %d.",
 			  ControlFile->toast_max_chunk_size, (int) TOAST_MAX_CHUNK_SIZE),
 				 errhint("It looks like you need to recompile or initdb.")));
+	if (ControlFile->loblksize != LOBLKSIZE)
+		ereport(FATAL,
+				(errmsg("database files are incompatible with server"),
+		  errdetail("The database cluster was initialized with LOBLKSIZE %d,"
+					" but the server was compiled with LOBLKSIZE %d.",
+					ControlFile->loblksize, (int) LOBLKSIZE),
+				 errhint("It looks like you need to recompile or initdb.")));
 
 #ifdef HAVE_INT64_TIMESTAMP
 	if (ControlFile->enableIntTimes != true)
diff --git a/src/backend/storage/large_object/inv_api.c b/src/backend/storage/large_object/inv_api.c
index 57ec1c2a6f9..09181423011 100644
--- a/src/backend/storage/large_object/inv_api.c
+++ b/src/backend/storage/large_object/inv_api.c
@@ -173,13 +173,38 @@ myLargeObjectExists(Oid loid, Snapshot snapshot)
 }
 
 
-static int32
-getbytealen(bytea *data)
+/*
+ * Extract data field from a pg_largeobject tuple, detoasting if needed
+ * and verifying that the length is sane.  Returns data pointer (a bytea *),
+ * data length, and an indication of whether to pfree the data pointer.
+ */
+static void
+getdatafield(Form_pg_largeobject tuple,
+			 bytea **pdatafield,
+			 int *plen,
+			 bool *pfreeit)
 {
-	Assert(!VARATT_IS_EXTENDED(data));
-	if (VARSIZE(data) < VARHDRSZ)
-		elog(ERROR, "invalid VARSIZE(data)");
-	return (VARSIZE(data) - VARHDRSZ);
+	bytea	   *datafield;
+	int			len;
+	bool		freeit;
+
+	datafield = &(tuple->data); /* see note at top of file */
+	freeit = false;
+	if (VARATT_IS_EXTENDED(datafield))
+	{
+		datafield = (bytea *)
+			heap_tuple_untoast_attr((struct varlena *) datafield);
+		freeit = true;
+	}
+	len = VARSIZE(datafield) - VARHDRSZ;
+	if (len < 0 || len > LOBLKSIZE)
+		ereport(ERROR,
+				(errcode(ERRCODE_DATA_CORRUPTED),
+				 errmsg("pg_largeobject entry for OID %u, page %d has invalid data field size %d",
+						tuple->loid, tuple->pageno, len)));
+	*pdatafield = datafield;
+	*plen = len;
+	*pfreeit = freeit;
 }
 
 
@@ -366,20 +391,14 @@ inv_getsize(LargeObjectDesc *obj_desc)
 	{
 		Form_pg_largeobject data;
 		bytea	   *datafield;
+		int			len;
 		bool		pfreeit;
 
 		if (HeapTupleHasNulls(tuple))	/* paranoia */
 			elog(ERROR, "null field found in pg_largeobject");
 		data = (Form_pg_largeobject) GETSTRUCT(tuple);
-		datafield = &(data->data);		/* see note at top of file */
-		pfreeit = false;
-		if (VARATT_IS_EXTENDED(datafield))
-		{
-			datafield = (bytea *)
-				heap_tuple_untoast_attr((struct varlena *) datafield);
-			pfreeit = true;
-		}
-		lastbyte = (uint64) data->pageno * LOBLKSIZE + getbytealen(datafield);
+		getdatafield(data, &datafield, &len, &pfreeit);
+		lastbyte = (uint64) data->pageno * LOBLKSIZE + len;
 		if (pfreeit)
 			pfree(datafield);
 	}
@@ -506,15 +525,7 @@ inv_read(LargeObjectDesc *obj_desc, char *buf, int nbytes)
 			off = (int) (obj_desc->offset - pageoff);
 			Assert(off >= 0 && off < LOBLKSIZE);
 
-			datafield = &(data->data);	/* see note at top of file */
-			pfreeit = false;
-			if (VARATT_IS_EXTENDED(datafield))
-			{
-				datafield = (bytea *)
-					heap_tuple_untoast_attr((struct varlena *) datafield);
-				pfreeit = true;
-			}
-			len = getbytealen(datafield);
+			getdatafield(data, &datafield, &len, &pfreeit);
 			if (len > off)
 			{
 				n = len - off;
@@ -630,16 +641,7 @@ inv_write(LargeObjectDesc *obj_desc, const char *buf, int nbytes)
 			 *
 			 * First, load old data into workbuf
 			 */
-			datafield = &(olddata->data);		/* see note at top of file */
-			pfreeit = false;
-			if (VARATT_IS_EXTENDED(datafield))
-			{
-				datafield = (bytea *)
-					heap_tuple_untoast_attr((struct varlena *) datafield);
-				pfreeit = true;
-			}
-			len = getbytealen(datafield);
-			Assert(len <= LOBLKSIZE);
+			getdatafield(olddata, &datafield, &len, &pfreeit);
 			memcpy(workb, VARDATA(datafield), len);
 			if (pfreeit)
 				pfree(datafield);
@@ -815,19 +817,11 @@ inv_truncate(LargeObjectDesc *obj_desc, int64 len)
 	if (olddata != NULL && olddata->pageno == pageno)
 	{
 		/* First, load old data into workbuf */
-		bytea	   *datafield = &(olddata->data);		/* see note at top of
-														 * file */
-		bool		pfreeit = false;
+		bytea	   *datafield;
 		int			pagelen;
+		bool		pfreeit;
 
-		if (VARATT_IS_EXTENDED(datafield))
-		{
-			datafield = (bytea *)
-				heap_tuple_untoast_attr((struct varlena *) datafield);
-			pfreeit = true;
-		}
-		pagelen = getbytealen(datafield);
-		Assert(pagelen <= LOBLKSIZE);
+		getdatafield(olddata, &datafield, &pagelen, &pfreeit);
 		memcpy(workb, VARDATA(datafield), pagelen);
 		if (pfreeit)
 			pfree(datafield);
diff --git a/src/bin/pg_controldata/pg_controldata.c b/src/bin/pg_controldata/pg_controldata.c
index 24e211668d1..f81502466e5 100644
--- a/src/bin/pg_controldata/pg_controldata.c
+++ b/src/bin/pg_controldata/pg_controldata.c
@@ -287,6 +287,8 @@ main(int argc, char *argv[])
 		   ControlFile.indexMaxKeys);
 	printf(_("Maximum size of a TOAST chunk:        %u\n"),
 		   ControlFile.toast_max_chunk_size);
+	printf(_("Size of a large-object chunk:         %u\n"),
+		   ControlFile.loblksize);
 	printf(_("Date/time type storage:               %s\n"),
 		   (ControlFile.enableIntTimes ? _("64-bit integers") : _("floating-point numbers")));
 	printf(_("Float4 argument passing:              %s\n"),
diff --git a/src/bin/pg_resetxlog/pg_resetxlog.c b/src/bin/pg_resetxlog/pg_resetxlog.c
index d11280e1d4d..915a1ed6463 100644
--- a/src/bin/pg_resetxlog/pg_resetxlog.c
+++ b/src/bin/pg_resetxlog/pg_resetxlog.c
@@ -52,6 +52,7 @@
 #include "catalog/catversion.h"
 #include "catalog/pg_control.h"
 #include "common/fe_memutils.h"
+#include "storage/large_object.h"
 #include "pg_getopt.h"
 
 
@@ -536,6 +537,7 @@ GuessControlValues(void)
 	ControlFile.nameDataLen = NAMEDATALEN;
 	ControlFile.indexMaxKeys = INDEX_MAX_KEYS;
 	ControlFile.toast_max_chunk_size = TOAST_MAX_CHUNK_SIZE;
+	ControlFile.loblksize = LOBLKSIZE;
 #ifdef HAVE_INT64_TIMESTAMP
 	ControlFile.enableIntTimes = true;
 #else
@@ -620,6 +622,8 @@ PrintControlValues(bool guessed)
 		   ControlFile.indexMaxKeys);
 	printf(_("Maximum size of a TOAST chunk:        %u\n"),
 		   ControlFile.toast_max_chunk_size);
+	printf(_("Size of a large-object chunk:         %u\n"),
+		   ControlFile.loblksize);
 	printf(_("Date/time type storage:               %s\n"),
 		   (ControlFile.enableIntTimes ? _("64-bit integers") : _("floating-point numbers")));
 	printf(_("Float4 argument passing:              %s\n"),
diff --git a/src/include/catalog/pg_control.h b/src/include/catalog/pg_control.h
index 4579eb6cef6..ba79d253ae5 100644
--- a/src/include/catalog/pg_control.h
+++ b/src/include/catalog/pg_control.h
@@ -21,7 +21,7 @@
 
 
 /* Version identifier for this pg_control format */
-#define PG_CONTROL_VERSION	941
+#define PG_CONTROL_VERSION	942
 
 /*
  * Body of CheckPoint XLOG records.  This is declared here because we keep
@@ -207,6 +207,7 @@ typedef struct ControlFileData
 	uint32		indexMaxKeys;	/* max number of columns in an index */
 
 	uint32		toast_max_chunk_size;	/* chunk size in TOAST tables */
+	uint32		loblksize;		/* chunk size in pg_largeobject */
 
 	/* flag indicating internal format of timestamp, interval, time */
 	bool		enableIntTimes; /* int64 storage enabled? */
diff --git a/src/include/storage/large_object.h b/src/include/storage/large_object.h
index 0d81a4bc1b6..30438a98cf3 100644
--- a/src/include/storage/large_object.h
+++ b/src/include/storage/large_object.h
@@ -66,6 +66,8 @@ typedef struct LargeObjectDesc
  * Also, it seems to be a smart move to make the page size be a power of 2,
  * since clients will often be written to send data in power-of-2 blocks.
  * This avoids unnecessary tuple updates caused by partial-page writes.
+ *
+ * NB: Changing LOBLKSIZE requires an initdb.
  */
 #define LOBLKSIZE		(BLCKSZ / 4)
 
-- 
GitLab