On 20/01/17 00:51, David Ahern wrote:
MPLS multipath for LSR is broken -- always selecting the first nexthop
in the one label case. For example:
$ ip netns exec ns1 ip -f mpls ro ls
100
nexthop as to 200 via inet 172.16.2.2 dev virt12
nexthop as to 300 via inet 172.16.3.2 dev virt13
101
nexthop as to 201 via inet6 2000:2::2 dev virt12
nexthop as to 301 via inet6 2000:3::2 dev virt13
In this example incoming packets have a single MPLS labels which means
BOS bit is set. The BOS bit is passed from mpls_forward down to
mpls_multipath_hash which never processes the hash loop because BOS is 1.
Removing the bos arg from mpls_multipath_hash uncovers a number of other
problems with the hash loop that processes the MPLS label stack -- from
incorrect assumptions on the skb (skb has already pulled the first mpls
label in mpls_forward yet loop assumes it is there)
This was intentional because it doesn't really add anything to include
the top-most label in the entropy since all traffic for the mpls_route
will have the same top-most label, until support for sharing of
mpls_routes is added.
Having said that, it costs very little to do this, makes the code
simpler and avoids the need to remember to change this if sharing is
added, so it's fine with me.
to incorrect
pskb_may_pull checks (label_index starts at 0 and pskb_may_pull checks
all use sizeof() * label_index).
This patch addresses all problems by moving the skb_pull in mpls_forward
after mpls_select_multipath. This allows mpls_multipath_hash to see the
skb with the entire label stack as it arrived.
From there mpls_multipath_hash is modified to additively compute the
total mpls header length on each pass (on pass N mpls_hdr_len is
N * sizeof(mpls_shim_hdr)). When the label is found with the BOS set it
verifies the skb has sufficient header for ipv4 or ipv6, and find the
IPv4 and IPv6 header by using the last mpls_hdr pointer and adding 1 to
advance past it.
With these changes I have verified the code correctly sees the label,
BOS, IPv4 and IPv6 addresses in the network header and icmp/tcp/udp
traffic for ipv4 and ipv6 are distributed across the nexthops.
Fixes: 1c78efa8319ca ("mpls: flow-based multipath selection")
Signed-off-by: David Ahern <d...@cumulusnetworks.com>
Acked-by: Robert Shearman <rshea...@brocade.com>
Good catch, thanks for fixing.
