kern/40621: net/if_srt.c uses wrong byte order for masks on little endian systems

[Wolfgang.Stukenbrock%nagler-company.com@localhost]

>Number:         40621
>Category:       kern
>Synopsis:       net/if_srt.c uses wrong byte order for masks on little endian 
>systems
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Feb 12 10:50:00 +0000 2009
>Originator:     Dr. Wolfgang Stukenbrock
>Release:        NetBSD 5_RC
>Organization:
Dr. Nagler & Company GmbH
        
>Environment:
        
        
System: NetBSD s012 4.0 NetBSD 4.0 (NSW-S012) #1: Thu Sep 11 12:21:03 CEST 2008 
root@s012:/usr/src/sys/arch/amd64/compile/NSW-S012 amd64
Architecture: x86_64
Machine: amd64
>Description:
        In order to forward a packet dependant on the source address, the 
if_srt.c interface will search a matching entry in a list of configured rules.
        During compare a static netmask-field is used.
        This field is initialized in a static way at compile time.
        The network-addresses are in network-byte-order when the compare is 
done, but the netmask-array is initialized with host-byte-order.
        The compare operation will not produce the indented results ...
>How-To-Repeat:
        Try to use the source-address-routing module on e.g. an i386-machine.
>Fix:
        apply the following fix to ths sourde file "src/sys/net/if_srt.c".
        The htonl() macro will resolve to something that may be uses in a 
static initialisation as long as the GNU-compiler is called with an 
optimisation flag set.
        (see /usr/include/sys/bswap.h) This is true for all cases - as long as 
I know - so it can be used here.
        This fix will also change the type of the array from "unsigned int" to 
"in_addr_t", because this is the type used for IPv4 addresses used for s_addr.

        remark: if for any reasons the use of htonl() should be avoided in the 
static initialisation, than the array should be initialized in host-byte-order
        as done before an a short loop should be added in srtattach() that will 
correct the byte-order once. From my point of view it should be no option to do
        the htonl() each time the compare is done for performance reasons.

--- if_srt.c    2009/02/12 09:58:04     1.1
+++ if_srt.c    2009/02/12 10:17:24
@@ -58,16 +58,17 @@
 
 /* Internal routines. */
 
-static unsigned int ipv4_masks[33]
- = { 0x00000000, /* /0 */
-     0x80000000, 0xc0000000, 0xe0000000, 0xf0000000, /* /1 - /4 */
-     0xf8000000, 0xfc000000, 0xfe000000, 0xff000000, /* /5 - /8 */
-     0xff800000, 0xffc00000, 0xffe00000, 0xfff00000, /* /9 - /12 */
-     0xfff80000, 0xfffc0000, 0xfffe0000, 0xffff0000, /* /13 - /16 */
-     0xffff8000, 0xffffc000, 0xffffe000, 0xfffff000, /* /17 - /20 */
-     0xfffff800, 0xfffffc00, 0xfffffe00, 0xffffff00, /* /21 - /24 */
-     0xffffff80, 0xffffffc0, 0xffffffe0, 0xfffffff0, /* /25 - /28 */
-     0xfffffff8, 0xfffffffc, 0xfffffffe, 0xffffffff  /* /29 - /32 */ };
+/* remark: in the NetBSD header fiels htonl() resolves to something usable in 
static initialisation */
+static const in_addr_t ipv4_masks[33]
+ = { htonl(0x00000000), /* /0 */
+     htonl(0x80000000), htonl(0xc0000000), htonl(0xe0000000), 
htonl(0xf0000000), /* /1 - /4 */
+     htonl(0xf8000000), htonl(0xfc000000), htonl(0xfe000000), 
htonl(0xff000000), /* /5 - /8 */
+     htonl(0xff800000), htonl(0xffc00000), htonl(0xffe00000), 
htonl(0xfff00000), /* /9 - /12 */
+     htonl(0xfff80000), htonl(0xfffc0000), htonl(0xfffe0000), 
htonl(0xffff0000), /* /13 - /16 */
+     htonl(0xffff8000), htonl(0xffffc000), htonl(0xffffe000), 
htonl(0xfffff000), /* /17 - /20 */
+     htonl(0xfffff800), htonl(0xfffffc00), htonl(0xfffffe00), 
htonl(0xffffff00), /* /21 - /24 */
+     htonl(0xffffff80), htonl(0xffffffc0), htonl(0xffffffe0), 
htonl(0xfffffff0), /* /25 - /28 */
+     htonl(0xfffffff8), htonl(0xfffffffc), htonl(0xfffffffe), 
htonl(0xffffffff)  /* /29 - /32 */ };
 
 static void update_mtu(SOFTC *sc)
 {

>Unformatted: