Skip to contentSkip to conversation

libupnp: add upstream security fix for CVE-2016-6255

PATCH New Peter Korsgaard 2 participants 4 messages Dec 19, 2016

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 libupnp: add upstream security fix for CVE-2016-6255

If there's no registered handler for a POST request, the default behaviour is to write it to the filesystem. Several million deployed devices appear to have this behaviour, making it possible to (at least) store arbitrary data on them. Add a configure option that enables this behaviour, and change the default to just drop POSTs that aren't directly handled.

---
 ...-unhandled-POSTs-to-write-to-the-filesyst.patch | 73 ++++++++++++++++++++++
 package/libupnp/libupnp.mk                         |  2 +
 2 files changed, 75 insertions(+)
 create mode 100644 package/libupnp/0001-Don-t-allow-unhandled-POSTs-to-write-to-the-filesyst.patch
package/libupnp/0001-Don-t-allow-unhandled-POSTs-to-write-to-the-filesyst.patch new file +73 -0
new file mode 100644
index 0000000..4bf1473
@@ -0,0 +1,73 @@
+From c91a8a3903367e1163765b73eb4d43be7d7927fa Mon Sep 17 00:00:00 2001
+From: Matthew Garrett <[email protected]>
+Date: Tue, 23 Feb 2016 13:53:20 -0800
+Subject: [PATCH] Don't allow unhandled POSTs to write to the filesystem by
+ default
+
+Fixes CVE-2016-6255: write files via POST
+
+If there's no registered handler for a POST request, the default behaviour
+is to write it to the filesystem. Several million deployed devices appear
+to have this behaviour, making it possible to (at least) store arbitrary
+data on them. Add a configure option that enables this behaviour, and change
+the default to just drop POSTs that aren't directly handled.
+
+Signed-off-by: Marcelo Roberto Jimenez <[email protected]>
+Signed-off-by: Peter Korsgaard <[email protected]>
+---
+ configure.ac                         | 4 ++++
+ upnp/inc/upnpconfig.h.in             | 5 +++++
+ upnp/src/genlib/net/http/webserver.c | 4 ++++
+ 3 files changed, 13 insertions(+)
+
+diff --git a/configure.ac b/configure.ac
+index dd88734..ea2bc09 100644
+--- a/configure.ac
++++ b/configure.ac
+@@ -482,6 +482,10 @@ if test "x$enable_scriptsupport" = xyes ; then
+         AC_DEFINE(IXML_HAVE_SCRIPTSUPPORT, 1, [see upnpconfig.h])
+ fi
+ 
++RT_BOOL_ARG_ENABLE([postwrite], [no], [write to the filesystem on otherwise unhandled POST requests])
++if test "x$enable_postwrite" = xyes ; then
++        AC_DEFINE(UPNP_ENABLE_POST_WRITE, 1, [see upnpconfig.h])
++fi
+ 
+ RT_BOOL_ARG_ENABLE([samples], [yes], [compilation of upnp/sample/ code])
+ 
+diff --git a/upnp/inc/upnpconfig.h.in b/upnp/inc/upnpconfig.h.in
+index 46ddc6e..5df8c5a 100644
+--- a/upnp/inc/upnpconfig.h.in
++++ b/upnp/inc/upnpconfig.h.in
+@@ -135,5 +135,10 @@
+  *  (i.e. configure --enable-open_ssl) */
+ #undef UPNP_ENABLE_OPEN_SSL
+ 
++/** Defined to 1 if the library has been compiled to support filesystem writes on POST
++ *  (i.e. configure --enable-postwrite) */
++#undef UPNP_ENABLE_POST_WRITE
++
++
+ #endif /* UPNP_CONFIG_H */
+ 
+diff --git a/upnp/src/genlib/net/http/webserver.c b/upnp/src/genlib/net/http/webserver.c
+index 8991c16..8b2ecf2 100644
+--- a/upnp/src/genlib/net/http/webserver.c
++++ b/upnp/src/genlib/net/http/webserver.c
+@@ -1369,9 +1369,13 @@ static int http_RecvPostMessage(
+ 		if (Fp == NULL)
+ 			return HTTP_INTERNAL_SERVER_ERROR;
+ 	} else {
++#ifdef UPNP_ENABLE_POST_WRITE
+ 		Fp = fopen(filename, "wb");
+ 		if (Fp == NULL)
+ 			return HTTP_UNAUTHORIZED;
++#else
++		return HTTP_NOT_FOUND;
++#endif
+ 	}
+ 	parser->position = POS_ENTITY;
+ 	do {
+-- 
+2.10.2
+
package/libupnp/libupnp.mk +2 -0
index e12f410..2a4e078 100644
@@ -11,5 +11,7 @@ LIBUPNP_CONF_ENV = ac_cv_lib_compat_ftime=no
 LIBUPNP_INSTALL_STAGING = YES
 LIBUPNP_LICENSE = BSD-3c
 LIBUPNP_LICENSE_FILES = LICENSE
+# configure.ac patched by 0001-Don-t-allow-unhandled-POSTs-to-write-to-the-filesyst.patch
+LIBUPNP_AUTORECONF = YES
 
 $(eval $(autotools-package))
Signature · 1 line
2.10.2
#1 lore raw

PATCH 2/2 libupnp: add upstream security fix for CVE-2016-8863

Fix out-of-bound access in create_url_list()

If there is an invalid URL in URLS->buf after a valid one, uri_parse is called with out pointing after the allocated memory. As uri_parse writes to *out before returning an error the loop in create_url_list must be stopped early to prevent an out-of-bound access.

---
 ...bound-access-in-create_url_list-CVE-2016-.patch | 64 ++++++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 package/libupnp/0002-Fix-out-of-bound-access-in-create_url_list-CVE-2016-.patch
package/libupnp/0002-Fix-out-of-bound-access-in-create_url_list-CVE-2016-.patch new file +64 -0
new file mode 100644
index 0000000..1b74d3e
@@ -0,0 +1,64 @@
+From 9c099c2923ab4d98530ab5204af1738be5bddba7 Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Uwe=20Kleine-K=C3=B6nig?= <[email protected]>
+Date: Thu, 8 Dec 2016 17:11:53 +0100
+Subject: [PATCH] Fix out-of-bound access in create_url_list() (CVE-2016-8863)
+
+If there is an invalid URL in URLS->buf after a valid one, uri_parse is
+called with out pointing after the allocated memory. As uri_parse writes
+to *out before returning an error the loop in create_url_list must be
+stopped early to prevent an out-of-bound access
+
+Bug: https://sourceforge.net/p/pupnp/bugs/133/
+Bug-CVE: https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2016-8863
+Bug-Debian: https://bugs.debian.org/842093
+Bug-Redhat: https://bugzilla.redhat.com/show_bug.cgi?id=1388771
+(cherry picked from commit a0f6e719bc03c4d2fe6a4a42ef6b8761446f520b)
+Signed-off-by: Peter Korsgaard <[email protected]>
+---
+ upnp/src/gena/gena_device.c | 17 ++++++++++++-----
+ 1 file changed, 12 insertions(+), 5 deletions(-)
+
+diff --git a/upnp/src/gena/gena_device.c b/upnp/src/gena/gena_device.c
+index fb04a29..245c56b 100644
+--- a/upnp/src/gena/gena_device.c
++++ b/upnp/src/gena/gena_device.c
+@@ -1113,7 +1113,7 @@ static int create_url_list(
+ 	/*! [out] . */
+ 	URL_list *out)
+ {
+-    size_t URLcount = 0;
++    size_t URLcount = 0, URLcount2 = 0;
+     size_t i;
+     int return_code = 0;
+     uri_type temp;
+@@ -1155,16 +1155,23 @@ static int create_url_list(
+         }
+         memcpy( out->URLs, URLS->buff, URLS->size );
+         out->URLs[URLS->size] = 0;
+-        URLcount = 0;
+         for( i = 0; i < URLS->size; i++ ) {
+             if( ( URLS->buff[i] == '<' ) && ( i + 1 < URLS->size ) ) {
+                 if( ( ( return_code =
+                         parse_uri( &out->URLs[i + 1], URLS->size - i + 1,
+-                                   &out->parsedURLs[URLcount] ) ) ==
++                                   &out->parsedURLs[URLcount2] ) ) ==
+                       HTTP_SUCCESS )
+-                    && ( out->parsedURLs[URLcount].hostport.text.size !=
++                    && ( out->parsedURLs[URLcount2].hostport.text.size !=
+                          0 ) ) {
+-                    URLcount++;
++                    URLcount2++;
++                    if (URLcount2 >= URLcount)
++                        /*
++                         * break early here in case there is a bogus URL that
++                         * was skipped above. This prevents to access
++                         * out->parsedURLs[URLcount] which is beyond the
++                         * allocation.
++                         */
++                        break;
+                 } else {
+                     if( return_code == UPNP_E_OUTOF_MEMORY ) {
+                         free( out->URLs );
+-- 
+2.10.2
+
Signature · 1 line
2.10.2
#2 lore raw

PATCH 1/2 libupnp: add upstream security fix for CVE-2016-6255

In reply to Peter Korsgaard

Hello,

Quoted 6 lines · Peter Korsgaard

If there's no registered handler for a POST request, the default behaviour is to write it to the filesystem. Several million deployed devices appear to have this behaviour, making it possible to (at least) store arbitrary data on them. Add a configure option that enables this behaviour, and change the default to just drop POSTs that aren't directly handled.

Quoted 6 lines
---
 ...-unhandled-POSTs-to-write-to-the-filesyst.patch | 73 ++++++++++++++++++++++
 package/libupnp/libupnp.mk                         |  2 +
 2 files changed, 75 insertions(+)
 create mode 100644 package/libupnp/0001-Don-t-allow-unhandled-POSTs-to-write-to-the-filesyst.patch

I've applied both to master, thanks!

I have to say that these security issues are terrible. The first one because the feature by itself is really silly and one may wonder why someone would implement such a feature in the first place. The second one because when you see what the URL parsing code looks like, no wonder why there are some security bugs in it...

Best regards,

Thomas

Signature · 3 lines
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
#3 lore raw

PATCH 1/2 libupnp: add upstream security fix for CVE-2016-6255

Quoted 17 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Thomas" == Thomas Petazzoni <[email protected]> writes:

Hello,

Quoted 6 lines · Peter Korsgaard

If there's no registered handler for a POST request, the default behaviour is to write it to the filesystem. Several million deployed devices appear to have this behaviour, making it possible to (at least) store arbitrary data on them. Add a configure option that enables this behaviour, and change the default to just drop POSTs that aren't directly handled.

Quoted 6 lines

---
...-unhandled-POSTs-to-write-to-the-filesyst.patch | 73 ++++++++++++++++++++++
package/libupnp/libupnp.mk | 2 +

2 files changed, 75 insertions(+)

create mode 100644 package/libupnp/0001-Don-t-allow-unhandled-POSTs-to-write-to-the-filesyst.patch

Quoted 7 lines

I've applied both to master, thanks!

I have to say that these security issues are terrible. The first one because the feature by itself is really silly and one may wonder why someone would implement such a feature in the first place. The second one because when you see what the URL parsing code looks like, no wonder why there are some security bugs in it...

Yeah, libupnp isn't really what I would best of class code :/

Thanks for applying the patches!

Signature · 1 line
Bye, Peter Korsgaard