From: Joerg Sonnenberger <joerg@bec.de>
Date: Thu, 1 Dec 2016 18:56:43 +0000 (+0100)
Subject: Reread the CAB header skipping the self-extracting binary code.
X-Git-Tag: v3.3.0~105
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=88eb9e1d73fef46f04677c25b1697b8e25777ed3;p=thirdparty%2Flibarchive.git

Reread the CAB header skipping the self-extracting binary code.

Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=15 as found
by the "OSS-Fuzz" project.
---

diff --git a/libarchive/archive_read_support_format_cab.c b/libarchive/archive_read_support_format_cab.c
index fc70684af..099f4a83d 100644
--- a/libarchive/archive_read_support_format_cab.c
+++ b/libarchive/archive_read_support_format_cab.c
@@ -645,12 +645,13 @@ cab_read_header(struct archive_read *a)
 	cab = (struct cab *)(a->format->data);
 	if (cab->found_header == 0 &&
 	    p[0] == 'M' && p[1] == 'Z') {
-		/* This is an executable?  Must be self-extracting... 	*/
+		/* This is an executable?  Must be self-extracting... */
 		err = cab_skip_sfx(a);
 		if (err < ARCHIVE_WARN)
 			return (err);
 
-		if ((p = __archive_read_ahead(a, sizeof(*p), NULL)) == NULL)
+		/* Re-read header after processing the SFX. */
+		if ((p = __archive_read_ahead(a, 42, NULL)) == NULL)
 			return (truncated_error(a));
 	}