Re: [PATCH net-next v7 1/3] net: gro: add {inner_}network_offset to napi_gro_cb

[Richard Gobert]

Paolo Abeni wrote:
> On Fri, 2024-04-12 at 17:55 +0200, Richard Gobert wrote:
>> This patch adds network_offset and inner_network_offset to napi_gro_cb, and
>> makes sure both are set correctly. In the common path there's only one
>> write (skb_gro_reset_offset).
>>
>> Signed-off-by: Richard Gobert <richardbgobert@xxxxxxxxx>
> 
> Does not apply cleanly to net-next. You have to wait until the net
> dependency is merged into net-next before posting.
> 
>> ---
>>  drivers/net/geneve.c           |  1 +
>>  drivers/net/vxlan/vxlan_core.c |  1 +
>>  include/net/gro.h              | 18 ++++++++++++++++--
>>  net/8021q/vlan_core.c          |  2 ++
>>  net/core/gro.c                 |  1 +
>>  net/ethernet/eth.c             |  1 +
>>  net/ipv4/af_inet.c             |  5 +----
>>  net/ipv4/gre_offload.c         |  1 +
>>  net/ipv6/ip6_offload.c         |  8 ++++----
>>  9 files changed, 28 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c
>> index 9c18a39b0d0c..a6256ea1f5bc 100644
>> --- a/drivers/net/geneve.c
>> +++ b/drivers/net/geneve.c
>> @@ -545,6 +545,7 @@ static struct sk_buff *geneve_gro_receive(struct sock *sk,
>>  	if (!ptype)
>>  		goto out;
>>  
>> +	NAPI_GRO_CB(skb)->inner_network_offset = hlen;
>>  	pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
>>  	flush = 0;
>>  
>> diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c
>> index 6fb182d9d6e7..9fb93c3953c1 100644
>> --- a/drivers/net/vxlan/vxlan_core.c
>> +++ b/drivers/net/vxlan/vxlan_core.c
>> @@ -754,6 +754,7 @@ static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
>>  
>>  	vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
>>  	if (vh) {
>> +		NAPI_GRO_CB(skb)->inner_network_offset = skb_gro_offset(skb);
>>  		if (!vxlan_parse_gpe_proto(vh, &protocol))
>>  			goto out;
>>  		ptype = gro_find_receive_by_type(protocol);
> 
> What about vxlan_gro_receive? and fou/gue?
> 

No need to write in fou/gue functions, as both functions call
{inet,inet6}_offloads, which means if there's an IP/IPv6 header after
fou/gue - ipip_gro_receive will be called (or ip6ip6_gro_receive, or
sit_ip6ip6_gro_receive, etc), in which inner_network_offset is written.

vxlan_gro_receive calls eth_gro_receive, in which inner_network_offset
is written as well.

> Side note: the latter apparently exist mainly to make UDP-related
> changes more difficult, can we deprecated them once for all?
> 
> Thank,
> 
> Paolo
>