Adopt the GNU convention for handling tar-archive members exceeding 8GB.

tglsfdc · tglsfdc · commit 00cdd83521cf · 2015-11-21T20:21:31.000-05:00
The POSIX standard for tar headers requires archive member sizes to be
printed in octal with at most 11 digits, limiting the representable file
size to 8GB.  However, GNU tar and apparently most other modern tars
support a convention in which oversized values can be stored in base-256,
allowing any practical file to be a tar member.  Adopt this convention
to remove two limitations:
* pg_dump with -Ft output format failed if the contents of any one table
exceeded 8GB.
* pg_basebackup failed if the data directory contained any file exceeding
8GB.  (This would be a fatal problem for installations configured with a
table segment size of 8GB or more, and it has also been seen to fail when
large core dump files exist in the data directory.)

File sizes under 8GB are still printed in octal, so that no compatibility
issues are created except in cases that would have failed entirely before.

In addition, this patch fixes several bugs in the same area:

* In 9.3 and later, we'd defined tarCreateHeader's file-size argument as
size_t, which meant that on 32-bit machines it would write a corrupt tar
header for file sizes between 4GB and 8GB, even though no error was raised.
This broke both "pg_dump -Ft" and pg_basebackup for such cases.

* pg_restore from a tar archive would fail on tables of size between 4GB
and 8GB, on machines where either "size_t" or "unsigned long" is 32 bits.
This happened even with an archive file not affected by the previous bug.

* pg_basebackup would fail if there were files of size between 4GB and 8GB,
even on 64-bit machines.

* In 9.3 and later, "pg_basebackup -Ft" failed entirely, for any file size,
on 64-bit big-endian machines.

In view of these potential data-loss bugs, back-patch to all supported
branches, even though removal of the documented 8GB limit might otherwise
be considered a new feature rather than a bug fix.
diff --git a/doc/src/sgml/ref/pg_dump.sgml b/doc/src/sgml/ref/pg_dump.sgml
@@ -272,12 +272,12 @@ PostgreSQL documentation
          <listitem>
           <para>
            Output a <command>tar</command>-format archive suitable for input
-           into <application>pg_restore</application>. The tar-format is
-           compatible with the directory-format; extracting a tar-format
+           into <application>pg_restore</application>. The tar format is
+           compatible with the directory format: extracting a tar-format
            archive produces a valid directory-format archive.
-           However, the tar-format does not support compression and has a
-           limit of 8 GB on the size of individual tables. Also, the relative
-           order of table data items cannot be changed during restore.
+           However, the tar format does not support compression. Also, when
+           using tar format the relative order of table data items cannot be
+           changed during restore.
           </para>
          </listitem>
         </varlistentry>
@@ -1140,15 +1140,6 @@ CREATE DATABASE foo WITH TEMPLATE template0;
    catalogs might be left in the wrong state.
   </para>
 
-  <para>
-   Members of tar archives are limited to a size less than 8 GB.
-   (This is an inherent limitation of the tar file format.)  Therefore
-   this format cannot be used if the textual representation of any one table
-   exceeds that size.  The total size of a tar archive and any of the
-   other output formats is not limited, except possibly by the
-   operating system.
-  </para>
-
   <para>
    The dump file produced by <application>pg_dump</application>
    does not contain the statistics used by the optimizer to make
diff --git a/src/backend/replication/basebackup.c b/src/backend/replication/basebackup.c
@@ -698,7 +698,7 @@ SendBackupHeader(List *tablespaces)
 		}
 		else
 		{
-			Size	len;
+			Size		len;
 
 			len = strlen(ti->oid);
 			pq_sendint(&buf, len, 4);
@@ -1131,13 +1131,6 @@ sendDir(char *path, int basepathlen, bool sizeonly, List *tablespaces,
  */
 
 
-/*
- * Maximum file size for a tar member: The limit inherent in the
- * format is 2^33-1 bytes (nearly 8 GB).  But we don't want to exceed
- * what we can represent in pgoff_t.
- */
-#define MAX_TAR_MEMBER_FILELEN (((int64) 1 << Min(33, sizeof(pgoff_t)*8 - 1)) - 1)
-
 /*
  * Given the member, write the TAR header & send the file.
  *
@@ -1166,15 +1159,6 @@ sendFile(char *readfilename, char *tarfilename, struct stat * statbuf,
 				 errmsg("could not open file \"%s\": %m", readfilename)));
 	}
 
-	/*
-	 * Some compilers will throw a warning knowing this test can never be true
-	 * because pgoff_t can't exceed the compared maximum on their platform.
-	 */
-	if (statbuf->st_size > MAX_TAR_MEMBER_FILELEN)
-		ereport(ERROR,
-				(errmsg("archive member \"%s\" too large for tar format",
-						tarfilename)));
-
 	_tarWriteHeader(tarfilename, NULL, statbuf);
 
 	while ((cnt = fread(buf, 1, Min(sizeof(buf), statbuf->st_size - len), fp)) > 0)
diff --git a/src/bin/pg_basebackup/pg_basebackup.c b/src/bin/pg_basebackup/pg_basebackup.c
@@ -781,7 +781,7 @@ ReceiveTarFile(PGconn *conn, PGresult *res, int rownum)
 	bool		in_tarhdr = true;
 	bool		skip_file = false;
 	size_t		tarhdrsz = 0;
-	size_t		filesz = 0;
+	pgoff_t		filesz = 0;
 
 #ifdef HAVE_LIBZ
 	gzFile		ztarfile = NULL;
@@ -1046,7 +1046,7 @@ ReceiveTarFile(PGconn *conn, PGresult *res, int rownum)
 
 						skip_file = (strcmp(&tarhdr[0], "recovery.conf") == 0);
 
-						sscanf(&tarhdr[124], "%11o", (unsigned int *) &filesz);
+						filesz = read_tar_number(&tarhdr[124], 12);
 
 						padding = ((filesz + 511) & ~511) - filesz;
 						filesz += padding;
@@ -1139,7 +1139,7 @@ ReceiveAndUnpackTarFile(PGconn *conn, PGresult *res, int rownum)
 	char		current_path[MAXPGPATH];
 	char		filename[MAXPGPATH];
 	const char *mapped_tblspc_path;
-	int			current_len_left;
+	pgoff_t		current_len_left = 0;
 	int			current_padding = 0;
 	bool		basetablespace;
 	char	   *copybuf = NULL;
@@ -1208,20 +1208,10 @@ ReceiveAndUnpackTarFile(PGconn *conn, PGresult *res, int rownum)
 			}
 			totaldone += 512;
 
-			if (sscanf(copybuf + 124, "%11o", &current_len_left) != 1)
-			{
-				fprintf(stderr, _("%s: could not parse file size\n"),
-						progname);
-				disconnect_and_exit(1);
-			}
+			current_len_left = read_tar_number(&copybuf[124], 12);
 
 			/* Set permissions on the file */
-			if (sscanf(&copybuf[100], "%07o ", &filemode) != 1)
-			{
-				fprintf(stderr, _("%s: could not parse file mode\n"),
-						progname);
-				disconnect_and_exit(1);
-			}
+			filemode = read_tar_number(&copybuf[100], 8);
 
 			/*
 			 * All files are padded up to 512 bytes
@@ -2180,7 +2170,7 @@ main(int argc, char **argv)
 	if (replication_slot && !streamwal)
 	{
 		fprintf(stderr,
-				_("%s: replication slots can only be used with WAL streaming\n"),
+			_("%s: replication slots can only be used with WAL streaming\n"),
 				progname);
 		fprintf(stderr, _("Try \"%s --help\" for more information.\n"),
 				progname);
diff --git a/src/bin/pg_dump/pg_backup_tar.c b/src/bin/pg_dump/pg_backup_tar.c
@@ -78,13 +78,6 @@ typedef struct
 	ArchiveHandle *AH;
 } TAR_MEMBER;
 
-/*
- * Maximum file size for a tar member: The limit inherent in the
- * format is 2^33-1 bytes (nearly 8 GB).  But we don't want to exceed
- * what we can represent in pgoff_t.
- */
-#define MAX_TAR_MEMBER_FILELEN (((int64) 1 << Min(33, sizeof(pgoff_t)*8 - 1)) - 1)
-
 typedef struct
 {
 	int			hasSeek;
@@ -1049,7 +1042,7 @@ isValidTarHeader(char *header)
 	int			sum;
 	int			chk = tarChecksum(header);
 
-	sscanf(&header[148], "%8o", &sum);
+	sum = read_tar_number(&header[148], 8);
 
 	if (sum != chk)
 		return false;
@@ -1091,13 +1084,6 @@ _tarAddFile(ArchiveHandle *AH, TAR_MEMBER *th)
 					  strerror(errno));
 	fseeko(tmp, 0, SEEK_SET);
 
-	/*
-	 * Some compilers will throw a warning knowing this test can never be true
-	 * because pgoff_t can't exceed the compared maximum on their platform.
-	 */
-	if (th->fileLen > MAX_TAR_MEMBER_FILELEN)
-		exit_horribly(modulename, "archive member too large for tar format\n");
-
 	_tarWriteHeader(th);
 
 	while ((cnt = fread(buf, 1, sizeof(buf), tmp)) > 0)
@@ -1222,11 +1208,10 @@ _tarGetHeader(ArchiveHandle *AH, TAR_MEMBER *th)
 {
 	lclContext *ctx = (lclContext *) AH->formatData;
 	char		h[512];
-	char		tag[100];
+	char		tag[100 + 1];
 	int			sum,
 				chk;
-	size_t		len;
-	unsigned long ullen;
+	pgoff_t		len;
 	pgoff_t		hPos;
 	bool		gotBlock = false;
 
@@ -1249,7 +1234,7 @@ _tarGetHeader(ArchiveHandle *AH, TAR_MEMBER *th)
 
 		/* Calc checksum */
 		chk = tarChecksum(h);
-		sscanf(&h[148], "%8o", &sum);
+		sum = read_tar_number(&h[148], 8);
 
 		/*
 		 * If the checksum failed, see if it is a null block. If so, silently
@@ -1272,27 +1257,31 @@ _tarGetHeader(ArchiveHandle *AH, TAR_MEMBER *th)
 		}
 	}
 
-	sscanf(&h[0], "%99s", tag);
-	sscanf(&h[124], "%12lo", &ullen);
-	len = (size_t) ullen;
+	/* Name field is 100 bytes, might not be null-terminated */
+	strlcpy(tag, &h[0], 100 + 1);
+
+	len = read_tar_number(&h[124], 12);
 
 	{
-		char		buf[100];
+		char		posbuf[32];
+		char		lenbuf[32];
 
-		snprintf(buf, sizeof(buf), INT64_FORMAT, (int64) hPos);
-		ahlog(AH, 3, "TOC Entry %s at %s (length %lu, checksum %d)\n",
-			  tag, buf, (unsigned long) len, sum);
+		snprintf(posbuf, sizeof(posbuf), UINT64_FORMAT, (uint64) hPos);
+		snprintf(lenbuf, sizeof(lenbuf), UINT64_FORMAT, (uint64) len);
+		ahlog(AH, 3, "TOC Entry %s at %s (length %s, checksum %d)\n",
+			  tag, posbuf, lenbuf, sum);
 	}
 
 	if (chk != sum)
 	{
-		char		buf[100];
+		char		posbuf[32];
 
-		snprintf(buf, sizeof(buf), INT64_FORMAT, (int64) ftello(ctx->tarFH));
+		snprintf(posbuf, sizeof(posbuf), UINT64_FORMAT,
+				 (uint64) ftello(ctx->tarFH));
 		exit_horribly(modulename,
 					  "corrupt tar header found in %s "
 					  "(expected %d, computed %d) file position %s\n",
-					  tag, sum, chk, buf);
+					  tag, sum, chk, posbuf);
 	}
 
 	th->targetFile = pg_strdup(tag);
@@ -1307,7 +1296,8 @@ _tarWriteHeader(TAR_MEMBER *th)
 {
 	char		h[512];
 
-	tarCreateHeader(h, th->targetFile, NULL, th->fileLen, 0600, 04000, 02000, time(NULL));
+	tarCreateHeader(h, th->targetFile, NULL, th->fileLen,
+					0600, 04000, 02000, time(NULL));
 
 	/* Now write the completed header. */
 	if (fwrite(h, 1, 512, th->tarFH) != 512)
diff --git a/src/include/pgtar.h b/src/include/pgtar.h
@@ -19,5 +19,7 @@ enum tarError
 	TAR_SYMLINK_TOO_LONG
 };
 
-extern enum tarError tarCreateHeader(char *h, const char *filename, const char *linktarget, size_t size, mode_t mode, uid_t uid, gid_t gid, time_t mtime);
+extern enum tarError tarCreateHeader(char *h, const char *filename, const char *linktarget,
+			  pgoff_t size, mode_t mode, uid_t uid, gid_t gid, time_t mtime);
+extern uint64 read_tar_number(const char *s, int len);
 extern int	tarChecksum(char *header);
diff --git a/src/port/tar.c b/src/port/tar.c
