On 2025/7/2 13:56, Morten Brørup wrote:
>> From: huangdengdui [mailto:huangdeng...@huawei.com]
>> Sent: Wednesday, 2 July 2025 04.38
>>
>> Hi everyone,
>>
>> The current rte_net_get_ptype() only supports parsing packets with  a
>> single 0x8100 VLAN tag and two 0x88a8 VLAN tags,
>> but does not support processing packets with two 0x8100 VLAN tags.
>>
>> Some network cards (e.g., hns3) do support parsing packets with two
>> 0x8100 VLAN tags.
>> Can this API be extended to support such packets?
> 
> It already does; the layers parameter must have RTE_PTYPE_INNER_L2_MASK set.

Yes, by setting RTE_PTYPE_INNER_L2_MASK, packets with two 0x8100 VLAN tags can 
be parsed,
but they will be recognized as tunnel packets even though they are not actually 
tunnel packets.

The current branch that parses the 0x8100 VLAN tag only handles a single VLAN 
tag[1][2].
Should we extend it to support parsing two or more 0x8100 VLAN tags?

[1]https://elixir.bootlin.com/dpdk/v25.07-rc2/source/lib/net/rte_net.c#L348
[2]https://elixir.bootlin.com/dpdk/v25.07-rc2/source/lib/net/rte_net.c#L500

> 
>>
>> I understand that two 0x8100 VLAN tag packets aren't commonly used, so
>> I'm uncertain whether adding support is necessary.
> 
> AFAIR, it's called "VLAN Stacking".
> Stacking two 0x8100 VLAN tags is commonly used.
> E.g. the majority of consumer fiber internet connections in Denmark use VLAN 
> Stacking (not QinQ) at the network owners' Points-of-Interconnect for the 
> service providers to identify individual homes.

Oh, I'm not very familiar with this. If stacking two 0x8100 VLAN tags is a 
commonly used, it would make sense for rte_net_get_ptype() to support parsing 
such packets.

> 
>>
>> Anyone is welcome to discuss this and share their opinions.
>>
>> Thanks,
>> /Dengdui

Reply via email to