Re: [PATCH net-next 6/8] tcp: Fix conditions to determine checksum offload

[Tom Herbert <tom@herbertland.com>]

On Mon, Dec 14, 2015 at 11:39 AM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> On Mon, 2015-12-14 at 11:19 -0800, Tom Herbert wrote:
>> In tcp_send_sendpage and tcp_sendmsg we check the route capabilities to
>> determine if checksum offload can be performed. This check currently
>> does not take the IP protocol into account for devices that advertise
>> only one of NETIF_F_IPV6_CSUM or NETIF_F_IP_CSUM. This patch adds a
>> function to check capabilities for checksum offload with a socket
>> called sk_check_csum_caps. This function checks for specific IPv4 or
>> IPv6 offload support based on the family of the socket.
>>
>> Signed-off-by: Tom Herbert <tom@herbertland.com>
>> ---
>>  include/net/sock.h | 9 +++++++++
>>  net/ipv4/tcp.c     | 4 ++--
>>  2 files changed, 11 insertions(+), 2 deletions(-)
>>
>> diff --git a/include/net/sock.h b/include/net/sock.h
>> index 0ca22b0..ab0269f 100644
>> --- a/include/net/sock.h
>> +++ b/include/net/sock.h
>> @@ -1791,6 +1791,15 @@ static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
>>       sk->sk_route_caps &= ~flags;
>>  }
>>
>> +static inline bool sk_check_csum_caps(struct sock *sk)
>> +{
>> +     return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
>> +            (sk->sk_family == PF_INET &&
>> +             (sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
>> +            (sk->sk_family == PF_INET6 &&
>> +             (sk->sk_route_caps & NETIF_F_IPV6_CSUM));
>> +}
>> +
>
> Not sure this handles ipv6_mapped correctly ?
>
It would be handled correctly, but might be suboptimal since we could
be missing opportunities to offload. The original code would assume
that all IPv6 can be offloaded with just NETIF_F_IP_CSUM which
presumably is also suboptimal because we're missing opportunity to do
copy-csum. I don't know which is better to optimize for...

> Since this is going to be called quite often, we might set a single bit
> in sk_setup_caps() or elsewhere and test it here as fast as possible.
>
>
>