generic: workaround arptables arpt_arp structure padding issue (#14577)

Linaro GCC 4.6-2013.05 incorrectly pads the arpt_arp structure on
at least the ar71xx target, resulting in a two bytes shorter struct
in the kernel compared to what arptables userspace expects.

When comparing sizeof(struct arpt_arp) in kernel and userspace, the
former yields 162 byte, while the latter is 164 byte.

As a consequence, the subsequent target_offset and next_offset
members of the parent arpt_entry structure contain invalid values
when processed by the arptables binary, leading to bad memory
accesses in the populate_cache() procedure, subsequently causing a
segfault.

Signed-off-by: Jo-Philipp Wich <[email protected]>

SVN-Revision: 38999

target/linux/generic/patches-3.10/614-netfilter_arpt_arp_padding_fix.patch

@@ -0,0 +1,16 @@
+--- a/include/uapi/linux/netfilter_arp/arp_tables.h
++++ b/include/uapi/linux/netfilter_arp/arp_tables.h
+@@ -68,6 +68,13 @@ struct arpt_arp {
+ 	__u8 flags;
+ 	/* Inverse flags */
+ 	__u16 invflags;
++
++	/* For some reason Linaro GCC 4.6-2013.05 omits the two
++	 * byte padding here when building for MIPS 34kc, leading to
++	 * userspace abi mismatches.
++	 * Add a dummy field to align this struct to 4 bytes
++	 */
++	__u16 padding;
+ };
+ 
+ /* Values for "flag" field in struct arpt_ip (general arp structure).