esp-hw-offload support for VF of NVIDIA Mellanox ConnectX-6 Ethernet Adapter Cards

esp-hw-offload support for VF of NVIDIA Mellanox ConnectX-6 Ethernet Adapter Cards

[高钧浩]

Hi Boris,Saeed

     I'm enabling esp-hw-offload for VF of NVIDIA Mellanox ConnectX-6
Ethernet Adapter Cards, but it doesn't work.
     Before I created VF, the esp-hw-offload function of CX-6 is on,
after I created VF, the esp-hw-offload function of VF doesn't inherit
the esp-hw-offload function of CX-6.
     Enable esp-hw-offload could refer to
https://docs.mellanox.com/display/OFEDv522200/IPsec+Crypto+Offload.

     Create VF steps as follows:
     modprobe mlx5_core
     echo 2 > /sys/class/net/net2/device/sriov_numvfs
     # lspci to get pci bdf number(example:0000:07:00.0)
     lspci -nn | grep Mellanox
     echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/unbind
     echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/unbind
     /etc/init.d/mst start
     mcra /dev/mst/mt4119_pciconf0  0x31500.17  0
     devlink dev eswitch set pci/0000:07:00.0  mode switchdev encap enable
     echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/bind
     echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/bind

     Then query the esp-hw-offload of VF:
     #firstly need to find the created VF(has the properties:
     bus-info: 0000:07:00.2, driver: mlx5_core)
     ethtool -i eth0 | grep esp-hw-offload
     esp-hw-offload: off [fixed]

Best Regards,
Junhao

Re: esp-hw-offload support for VF of NVIDIA Mellanox ConnectX-6 Ethernet Adapter Cards

[Saeed Mahameed]

On Mon, 2021-03-29 at 16:42 +0800, 高钧浩 wrote:
> Hi Boris,Saeed
> 
>      I'm enabling esp-hw-offload for VF of NVIDIA Mellanox ConnectX-6
> Ethernet Adapter Cards, but it doesn't work.
>      Before I created VF, the esp-hw-offload function of CX-6 is on,
> after I created VF, the esp-hw-offload function of VF doesn't inherit
> the esp-hw-offload function of CX-6.
>      Enable esp-hw-offload could refer to
> https://docs.mellanox.com/display/OFEDv522200/IPsec+Crypto+Offload.
> 
>      Create VF steps as follows:
>      modprobe mlx5_core
>      echo 2 > /sys/class/net/net2/device/sriov_numvfs
>      # lspci to get pci bdf number(example:0000:07:00.0)
>      lspci -nn | grep Mellanox
>      echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/unbind
>      echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/unbind
>      /etc/init.d/mst start
>      mcra /dev/mst/mt4119_pciconf0  0x31500.17  0
>      devlink dev eswitch set pci/0000:07:00.0  mode switchdev encap
> enable
>      echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/bind
>      echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/bind
> 
>      Then query the esp-hw-offload of VF:
>      #firstly need to find the created VF(has the properties:
>      bus-info: 0000:07:00.2, driver: mlx5_core)
>      ethtool -i eth0 | grep esp-hw-offload
>      esp-hw-offload: off [fixed]
> 

Huy, Raed, Do you know if we support IPsec inline offload on VFs ?

Re: esp-hw-offload support for VF of NVIDIA Mellanox ConnectX-6 Ethernet Adapter Cards

[高钧浩]

Ping ...

Saeed Mahameed <saeed@kernel.org> 于2021年3月30日周二 上午5:05写道：
>
> On Mon, 2021-03-29 at 16:42 +0800, 高钧浩 wrote:
> > Hi Boris,Saeed
> >
> >      I'm enabling esp-hw-offload for VF of NVIDIA Mellanox ConnectX-6
> > Ethernet Adapter Cards, but it doesn't work.
> >      Before I created VF, the esp-hw-offload function of CX-6 is on,
> > after I created VF, the esp-hw-offload function of VF doesn't inherit
> > the esp-hw-offload function of CX-6.
> >      Enable esp-hw-offload could refer to
> > https://docs.mellanox.com/display/OFEDv522200/IPsec+Crypto+Offload.
> >
> >      Create VF steps as follows:
> >      modprobe mlx5_core
> >      echo 2 > /sys/class/net/net2/device/sriov_numvfs
> >      # lspci to get pci bdf number(example:0000:07:00.0)
> >      lspci -nn | grep Mellanox
> >      echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/unbind
> >      echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/unbind
> >      /etc/init.d/mst start
> >      mcra /dev/mst/mt4119_pciconf0  0x31500.17  0
> >      devlink dev eswitch set pci/0000:07:00.0  mode switchdev encap
> > enable
> >      echo 0000:07:00.2 > /sys/bus/pci/drivers/mlx5_core/bind
> >      echo 0000:07:00.3 > /sys/bus/pci/drivers/mlx5_core/bind
> >
> >      Then query the esp-hw-offload of VF:
> >      #firstly need to find the created VF(has the properties:
> >      bus-info: 0000:07:00.2, driver: mlx5_core)
> >      ethtool -i eth0 | grep esp-hw-offload
> >      esp-hw-offload: off [fixed]
> >
>
> Huy, Raed, Do you know if we support IPsec inline offload on VFs ?
>
>