[dpdk-dev] KNI PCIe number and VLAN Stripping

Ferruh Yigit ferruh.yigit at intel.com
Tue Mar 20 14:56:23 CET 2018


On 3/19/2018 1:10 AM, Yasuhiro Ohara wrote:
> 
> Hi.
> 
> It is much appreciated if someone could answer the following questions.
> 
> 1. Regarding KNI, what is the reasoning of giving the PCI device
> number information when setting kni I/F ? Can the data be blank or
> dummy ? I would like to create a virtual I/F inside my application
> and its corresponding kni, which may not have the corresponding
> physical PCIe slot. 

For KNI pci address information is for ethtool support [1] and it can be blank,
for example KNI PMD uses it blank.

[1]
ethtool support is only for igb and ixgbe only. And this support is by default
disabled in config, need to enable CONFIG_RTE_KNI_KMOD_ETHTOOL config option for it.

> 
> 2. the ETHER/VLAN/ARP packet looks not-supported both
> in mbuf_ptype(RTE_PTYPE_XXX in m->packet_type) and
> in PKT_RX_VLAN_STRIPPED situation.
>   2a. How can we encode in m->packet_type to indicate there are
>       arp_hdr after the (ether_hdr,vlan_hdr) pair ?
>   2b. How can we get the next header value (i.e. ether_proto in
>       vlan_hdr) when the vlan_hdr is stripped by the NIC ?
> 
> Thanks in advance.
> 
> Best regards,
> Yasu
> 



More information about the dev mailing list