Re: [PATCH net-next] gso: enable udp gso for virtual devices

[Jason Baron <jbaron@akamai.com>]

On 6/13/19 5:20 PM, Willem de Bruijn wrote:
>>>> @@ -237,6 +237,7 @@ static inline int find_next_netdev_feature(u64 feature, unsigned long start)
>>>>                                  NETIF_F_GSO_GRE_CSUM |                 \
>>>>                                  NETIF_F_GSO_IPXIP4 |                   \
>>>>                                  NETIF_F_GSO_IPXIP6 |                   \
>>>> +                                NETIF_F_GSO_UDP_L4 |                   \
>>>>                                  NETIF_F_GSO_UDP_TUNNEL |               \
>>>>                                  NETIF_F_GSO_UDP_TUNNEL_CSUM)
>>>
>>> Are you adding this to NETIF_F_GSO_ENCAP_ALL? Wouldn't it make more
>>> sense to add it to NETIF_F_GSO_SOFTWARE?
>>>
>>
>> Yes, I'm adding to NETIF_F_GSO_ENCAP_ALL (not very clear from the
>> context). I will fix the commit log.
>>
>> In: 83aa025 udp: add gso support to virtual devices, the support was
>> also added to NETIF_F_GSO_ENCAP_ALL (although subsequently reverted due
>> to UDP GRO not being in place), so I wonder what the reason was for that?
> 
> That was probably just a bad choice on my part.
> 
> It worked in practice, but if NETIF_F_GSO_SOFTWARE works the same
> without unexpected side effects, then I agree that it is the better choice.
> 
> That choice does appear to change behavior when sending over tunnel
> devices. Might it send tunneled GSO packets over loopback?
> 
> 

I set up a test case using fou tunneling through a bridge device using
the udpgso_bench_tx test where packets are not received correctly if
NETIF_F_GSO_UDP_L4 is added to NETIF_F_GSO_SOFTWARE. If I have it added
to NETIF_F_GSO_ENCAP_ALL, it does work correctly. So there are more
fixes required to include it in NETIF_F_GSO_SOFTWARE.

The use-case I have only requires it to be in NETIF_F_GSO_ENCAP_ALL, but
if it needs to go in NETIF_F_GSO_SOFTWARE, I can look at what's required
more next week.

Thanks,

-Jason

> 
>> I agree that NETIF_F_GSO_SOFTWARE seems conceptually more logical and
>> further I think it adds support for more 'virtual' devices. For example,
>> I tested loopback with NETIF_F_GSO_UDP_L4 being added to
>> NETIF_F_GSO_SOFTWARE and it shows a nice performance gain, whereas
>> NETIF_F_GSO_ENCAP_ALL isn't included for loopback.
>>
>> Thanks,
>>
>> -Jason