[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <e5c28c89-0022-69d2-8538-74c2c842a2fc@gmail.com>
Date: Wed, 23 Sep 2020 08:36:50 -0600
From: David Ahern <dsahern@...il.com>
To: Eric Dumazet <eric.dumazet@...il.com>,
Maciej Żenczykowski <zenczykowski@...il.com>,
Maciej Żenczykowski <maze@...gle.com>,
"David S . Miller" <davem@...emloft.net>
Cc: Linux Network Development Mailing List <netdev@...r.kernel.org>,
Willem de Bruijn <willemb@...gle.com>,
Lorenzo Colitti <lorenzo@...gle.com>,
Sunmeet Gill <sgill@...cinc.com>,
Vinay Paradkar <vparadka@....qualcomm.com>,
Tyler Wear <twear@...cinc.com>,
David Ahern <dsahern@...nel.org>
Subject: Re: [PATCH v2] net/ipv4: always honour route mtu during forwarding
On 9/23/20 2:46 AM, Eric Dumazet wrote:
>> diff --git a/include/net/ip.h b/include/net/ip.h
>> index b09c48d862cc..c2188bebbc54 100644
>> --- a/include/net/ip.h
>> +++ b/include/net/ip.h
>> @@ -436,12 +436,17 @@ static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
>> bool forwarding)
>> {
>> struct net *net = dev_net(dst->dev);
>> + unsigned int mtu;
>>
>> if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
>> ip_mtu_locked(dst) ||
>> !forwarding)
>> return dst_mtu(dst);
>>
>> + /* 'forwarding = true' case should always honour route mtu */
>> + mtu = dst_metric_raw(dst, RTAX_MTU);
>> + if (mtu) return mtu;
>
>
> if (mtu)
> return mtu;
>
> Apparently route mtu are capped to 65520, not sure where it is done exactly IP_MAX_MTU being 65535)
ip_metrics_convert seems to be the place"
if (type == RTAX_MTU && val > 65535 - 15)
val = 65535 - 15;
going back through the code moves, and it was added by Dave with
710ab6c03122c
>
> # ip ro add 1.1.1.4 dev wlp2s0 mtu 100000
> # ip ro get 1.1.1.4
> 1.1.1.4 dev wlp2s0 src 192.168.8.147 uid 0
> cache mtu 65520
>
>
>
>
>> +
>> return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
>> }
>>
>>
Powered by blists - more mailing lists