On Thu, 20 Feb 2014 11:28:49 +0900 Simon Horman <[email protected]> wrote:
> In my reading there default values that an MPLS LSE should take > on MPLS push is not entirely clear in some circumstances. > > This patch illustrates some discrepancies in the Ryu and Open vSwtich > implementations by moving the Ryu tests to match that of Open vSwtich. > > * When pushing an MPLS LSE onto a packet which is neither IP nor MPLS > use 64 as the TTL of the new LSE. > > Ryu currently uses 0. > > * When pushing an MPLS LSE onto an IPv6 packet > use 2 (IPv6 Explicit NULL) as the default label. > > Ryu currently uses 0 (IPv4 Explicit NULL). > > Another approach, which would neatly bypass the lack of clarity in > the spec would be to perform explicit set_mpls_ttl and set_label > actions as part of the tests. OF 1.3.3 spec (5.12.1) says: When executing a push action, for all elds specied in Table 8 part of the added header, the value from the corresponding eld in the existing outer headers of the packet should be copied in the new eld. If the corresponding outer header eld does not exist in the packet, the new field should be set to zero. = I understand the above in the following way: The default label value should be zero for this test (since no outer MLS header exists). The default TTL should be copied from an IP header. The ARP packet doesn't have an IP header so the default TTL should be zero. ------------------------------------------------------------------------------ Managing the Performance of Cloud-Based Applications Take advantage of what the Cloud has to offer - Avoid Common Pitfalls. Read the Whitepaper. http://pubads.g.doubleclick.net/gampad/clk?id=121054471&iu=/4140/ostg.clktrk _______________________________________________ Ryu-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ryu-devel
