On 11/18/20 3:34 AM, Ajit Khaparde wrote:
> On Mon, Nov 16, 2020 at 8:13 AM Ferruh Yigit <ferruh.yi...@intel.com> wrote:
>
>> On 11/16/2020 7:55 AM, Xiaoyu Min wrote:
>>> From: Xiaoyu Min <jack...@nvidia.com>
>>>
>>> The rte_flow_item_eth and rte_flow_item_vlan items are refined.
>>> The structs do not exactly represent the packet bits captured on the
>>> wire anymore so should only copy real header instead of the whole struct.
>>>
>>> Replace the rte_flow_item_* with the existing corresponding rte_*_hdr.
>>>
>>> Fixes: 09315fc83861 ("ethdev: add VLAN attributes to ethernet and VLAN
>> items")
>>>
>>> Signed-off-by: Xiaoyu Min <jack...@nvidia.com>
>>
>> <...>
>>
>>> @@ -1726,7 +1727,7 @@ ulp_rte_vxlan_encap_act_handler(const struct
>> rte_flow_action *action_item,
>>> BNXT_TF_DBG(ERR, "vxlan encap does not have vni\n");
>>> return BNXT_TF_RC_ERROR;
>>> }
>>> - vxlan_size = sizeof(struct rte_flow_item_vxlan);
>>> + vxlan_size = sizeof(struct rte_vxlan_hdr);
>>> /* copy the vxlan details */
>>> memcpy(&vxlan_spec, item->spec, vxlan_size);
>>> vxlan_spec.flags = 0x08;
>>>
>>
>> 'vxlan_size' seems used both to copy rt_flow_item [1] and to header [2].
>> Also
>> ''vxlan_size' is used to copy the 'vxlan_spec'.
>>
>> Since both "struct rte_flow_item_vxlan" & "struct rte_vxlan_hdr" size is
>> same,
>> this should work fine, but I guess it may be broken if sizes of those two
>> structures changes in the future.
>>
>> [1]
>> memcpy(&vxlan_spec, item->spec, vxlan_size);
>>
>> [2]
>> ulp_encap_buffer_copy(buff, (const uint8_t *)&vxlan_spec,
>> vxlan_size, ULP_BUFFER_ALIGN_8_BYTE);
>>
> Also, I feel rte_flow_item_vxlan is a control plane structure while
> rte_vlan_hdr is more for dataplane.
> It's better to keep them separate to allow refining them independently.
I disagree with the point. It is typical duplication which
should be fixed. VXLAN item should consist of rte_vxlan_hdr
plus extra non header fields. In fact, similar thing should
be done for Ethernet flow item spec.
These items are used for VXLAN encap action and in current
state of definitions it is tricky to use VXLAN and Ethernet
items to construct header to be prepended. It is required
to copy field by field since it is bad to rely on the fact
that flow item starts from protocol header if it is not
specified in corresponding C type definition.
