Re: [PATCH v2 net] gro_cells: Avoid packet re-ordering for cloned skbs

[Thomas Bogendoerfer]

On Wed, 29 Jan 2025 13:06:49 +0100
Eric Dumazet <edumazet@xxxxxxxxxx> wrote:

> "TCP Sender in namespace A -> ip6_tunnel -> ipvlan -> ipvlan ->
> ip6_tunnel -> TCP receiver"
> or
> " TCP Sender -> ip6_tunnel -> ipvlan -> ipvlan -> ip6_tunnel -> TCP Receiver"
> 
> In this case, GRO in ip6_tunnel is not needed at all, since proper TSO
> packets should already be cooked by TCP sender and be carried
> to the receiver as plain GRO packets.
> 
> gro_cells was added at a time GRO layer was only  supporting native
> encapsulations : IPv4 + TCP or IPv6 + TCP.
> 
> Nowadays, GRO supports encapsulated traffic just fine, same for TSO
> packets encapsulated in ip6_tunnel
> 
> Maybe it is time to remove gro_cells from net/ipv6/ip6_tunnel.c
> 
> diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
> index 48fd53b9897265338086136e96ea8e8c6ec3cac..b91c253dc4f1998f8df74251a93e29d00c03db5
> 100644
> --- a/net/ipv6/ip6_tunnel.c
> +++ b/net/ipv6/ip6_tunnel.c
> [...]

this patch works for my test case. So the same thing should be probably
done for net/ipv4/ip_tunnel.c and net/ipv6/ip6_gre.c, too ?

Thomas.

-- 
SUSE Software Solutions Germany GmbH
HRB 36809 (AG Nürnberg)
Geschäftsführer: Ivo Totev, Andrew McDonald, Werner Knoblich