Re: [PATCH v3 3/3] i40e: Add UDP segmentation offload support

["Samudrala, Sridhar" <sridhar.samudrala@intel.com>]

On 10/11/2019 10:56 AM, Alexander Duyck wrote:
> On Fri, 2019-10-11 at 10:39 -0700, Samudrala, Sridhar wrote:
>>
>> On 10/11/2019 9:53 AM, Josh Hunt wrote:
>>> Based on a series from Alexander Duyck this change adds UDP segmentation
>>> offload support to the i40e driver.
>>>
>>> CC: Alexander Duyck <alexander.h.duyck@intel.com>
>>> CC: Willem de Bruijn <willemb@google.com>
>>> Signed-off-by: Josh Hunt <johunt@akamai.com>
>>> ---
>>>    drivers/net/ethernet/intel/i40e/i40e_main.c |  1 +
>>>    drivers/net/ethernet/intel/i40e/i40e_txrx.c | 12 +++++++++---
>>>    2 files changed, 10 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
>>> index 6031223eafab..56f8c52cbba1 100644
>>> --- a/drivers/net/ethernet/intel/i40e/i40e_main.c
>>> +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
>>> @@ -12911,6 +12911,7 @@ static int i40e_config_netdev(struct i40e_vsi *vsi)
>>>    			  NETIF_F_GSO_IPXIP6		|
>>>    			  NETIF_F_GSO_UDP_TUNNEL	|
>>>    			  NETIF_F_GSO_UDP_TUNNEL_CSUM	|
>>> +			  NETIF_F_GSO_UDP_L4		|
>>>    			  NETIF_F_SCTP_CRC		|
>>>    			  NETIF_F_RXHASH		|
>>>    			  NETIF_F_RXCSUM		|
>>> diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
>>> index e3f29dc8b290..b8496037ef7f 100644
>>> --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
>>> +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
>>> @@ -2960,10 +2960,16 @@ static int i40e_tso(struct i40e_tx_buffer *first, u8 *hdr_len,
>>>    
>>>    	/* remove payload length from inner checksum */
>>>    	paylen = skb->len - l4_offset;
>>> -	csum_replace_by_diff(&l4.tcp->check, (__force __wsum)htonl(paylen));
>>>    
>>> -	/* compute length of segmentation header */
>>> -	*hdr_len = (l4.tcp->doff * 4) + l4_offset;
>>> +	if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) {
>>> +		csum_replace_by_diff(&l4.udp->check, (__force __wsum)htonl(paylen));
>>> +		/* compute length of segmentation header */
>>> +		*hdr_len = sizeof(*l4.udp) + l4_offset;
>>> +	} else {
>>> +		csum_replace_by_diff(&l4.tcp->check, (__force __wsum)htonl(paylen));
>>> +		/* compute length of segmentation header */
>>> +		*hdr_len = (l4.tcp->doff * 4) + l4_offset;
>>> +	}
>>
>> Is it guaranteed that gso_type can be either UDP or TCP only if we reach
>> here? Don't we need to handle the case where it is neither and return
>> from this function?
> 
> We should only reach here if a supported gso_type value is in the packet,
> otherwise we should end up with software segmentation taking care of it
> and clearing the gso_size value if I recall correctly.
> 
> Otherwise the code should have been checking for non-TCP types ages ago,
> and we would have experienced all sorts of bugs.

Yes. Sounds good. Thanks for the clarification.

> 
> - Alex
>