On 3/27/19 3:08 AM, Ido Schimmel wrote:
> On Tue, Mar 26, 2019 at 08:29:36PM -0700, David Ahern wrote:
>> @@ -989,8 +989,11 @@ static void ip6_rt_copy_init(struct rt6_info *rt,
>> struct fib6_info *ort)
>>
>> rt->rt6i_dst = ort->fib6_dst;
>> rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
>> - rt->rt6i_gateway = ort->fib6_nh.nh_gw;
>> rt->rt6i_flags = ort->fib6_flags;
>> + if (ort->fib6_nh.fib_nh_has_gw) {
>> + rt->rt6i_gateway = ort->fib6_nh.nh_gw;
>> + rt->rt6i_flags |= RTF_GATEWAY;
>
> Not sure I understand why this line is needed. Can you explain?
The RTF_GATEWAY flag in an rt6_info still has many users.
>
>> + }
>> rt6_set_from(rt, ort);
>> #ifdef CONFIG_IPV6_SUBTREES
>> rt->rt6i_src = ort->fib6_src;
>> @@ -1870,7 +1873,7 @@ struct rt6_info *ip6_pol_route(struct net *net, struct
>> fib6_table *table,
>
> ...
>
>> @@ -3134,7 +3138,7 @@ static struct fib6_info *ip6_route_info_create(struct
>> fib6_config *cfg,
>> } else
>> rt->fib6_prefsrc.plen = 0;
>>
>> - rt->fib6_flags = cfg->fc_flags;
>> + rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
>
> This is going to break mlxsw. See:
>
> mlxsw_sp_fib6_rt_can_mp()
> mlxsw_sp_rt6_is_gateway()
hmmm... not sure how I missed those. Will rescan the code for any other
missed checks on a fib6_info.
