Re: [PATCH] net/ipv6: repeat route lookup with saddr set for ECMP

[Paolo Abeni]

On 3/29/26 11:12 AM, Maximilian Moehl wrote:
> When the routing decision involves ECMP, the initial hash is calculated
> with saddr being :: and the decision which interface to select from an
> ECMP group is based on that hash. If the route lookup has to be
> repeated, e.g. because a route was updated, the hash is calculated with
> saddr set to the previously selected address. This can cause the
> selected interface to change, breaking ongoing connections.
> 
> To ensure the initial interface selection is based on an actual saddr
> the route lookup is repeated after the source address selection if a
> hash was calculated.
> 
> Signed-off-by: Maximilian Moehl <maximilian@xxxxxxxx>
> Link: https://lore.kernel.org/all/aOYLRyIlc7XU7-7n@shredder/
> ---
> I've created a write-up of what I've done, including the steps taken
> to test the patch: https://moehl.eu/blog/linux-ipv6-ecmp-instability.html
> 
>  net/ipv6/ip6_output.c | 12 ++++++++++++
>  1 file changed, 12 insertions(+)
> 
> diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
> index 8e2a6b28cea7..465fce51d017 100644
> --- a/net/ipv6/ip6_output.c
> +++ b/net/ipv6/ip6_output.c
> @@ -1148,6 +1148,18 @@ static int ip6_dst_lookup_tail(struct net *net, const struct sock *sk,
>  			*dst = NULL;
>  		}
>  
> +		/* If ECMP was involved the initial hash was calculted
> +		 * with saddr=:: which can result in instability
> +		 * when it is later re-calculated with the selected
> +		 * saddr. Lookup the route again with the chosen
> +		 * saddr to get a stable result.
> +		 */
> +		if (fl6->mp_hash) {
> +			fl6->mp_hash = 0;
> +			dst_release(*dst);
> +			*dst = NULL;
> +		}
> +
>  		if (fl6->flowi6_oif)
>  			flags |= RT6_LOOKUP_F_IFACE;
>  	}

This apparently breaks ipv6 fib tests (fib_tests.sh):

# IPv6 multipath load balance test
#      TEST: IPv6 multipath loadbalance                          [FAIL]

see
https://github.com/linux-netdev/nipa/wiki/How-to-run-netdev-selftests-CI-style
on how to reproduce the tests.

Also this would deserve additional testcases.

Without diving much inside the code I have the feeling this change is
plugged into the wrong place: multipath selection logic should be
encapsulated by fib6_select_path().

/P