[dpdk-users] Cannot run DPDK applications using Mellanox NIC under Hyper-V

[Liwu Liu]

Hi Stephen,

  Yeah we are seeing all NETVSC, failsafe PMD, TAP, MLX4 VF. We are expecting bifurcation mode where default traffic goes with MLX4_EN slave to NETVSC master to enter Linux net stack and certain flow-defined traffic shall go into DPDK RX queue rather than MLX4_EN net-dev queue. Unfortunately as we cannot define flow into VF NIC, we cannot re-direct traffic into DPDK user-land. 

   In fact I am looking for a way to steer all RX traffic from MLX4_EN slave device into DPDK RX ring. 

  Hypervisor: Windows server + MLNX VPI 5.5; VM: Linux dpdk-18.11 mlnx-ofed-kernel-4.4 rdma-core-43mlnx1 

  Ethernet controller: Mellanox Technologies MT27500 Family [ConnectX-3 Virtual Function]

Best,

Liwu

-----Original Message-----
From: Stephen Hemminger <stephen at networkplumber.org> 
Sent: Friday, January 4, 2019 4:24 PM
To: Liwu Liu <liwuliu at fortinet.com>
Subject: Re: [dpdk-users] Cannot run DPDK applications using Mellanox NIC under Hyper-V

On Sat, 5 Jan 2019 00:10:43 +0000
Liwu Liu <liwuliu at fortinet.com> wrote:

> Hi Stephen,
> 
>    It is SR-IOV
> 
>   Thanks,
> 
> Liwu
> 
> -----Original Message-----
> From: Stephen Hemminger <stephen at networkplumber.org>
> Sent: Friday, January 4, 2019 4:07 PM
> To: Liwu Liu <liwuliu at fortinet.com>
> Cc: users at dpdk.org
> Subject: Re: [dpdk-users] Cannot run DPDK applications using Mellanox 
> NIC under Hyper-V
> 
> On Fri, 4 Jan 2019 20:06:48 +0000
> Liwu Liu <liwuliu at fortinet.com> wrote:
> 
> > Hi Team,
> >     We used to have similar problem for Ubuntu 18.04 hypervisor and 
> > resolved by set log_num_mgm_entry_size=-1. (Refer to 
> > https://mails.dpdk.org/archives/users/2018-November/003647.html
> > )
> > 
> >   Somehow for Windows Servers with MLNX VPI, I do not know where to set such and DPDK over MLX4 on linux vm has same failure of attaching flow.
> > 
> >     [  374.568992] <mlx4_ib> __mlx4_ib_create_flow: mcg table is 
> > full. Fail to register network rule. size = 64 (out of memory error 
> > code)
> > 
> >     Would like to get your help on this. It seems to be the case that the PF interface is not configured to trust VF interfaces to be able to add new flow rules.
> > 
> >    Many thanks for help,
> > 
> > Liwu
> > 
> > 
> > 
> > 
> > ***  Please note that this message and any attachments may contain 
> > confidential and proprietary material and information and are 
> > intended only for the use of the intended recipient(s). If you are 
> > not the intended recipient, you are hereby notified that any review, 
> > use, disclosure, dissemination, distribution or copying of this 
> > message and any attachments is strictly prohibited. If you have 
> > received this email in error, please immediately notify the sender 
> > and destroy this e-mail and any attachments and all copies, whether 
> > electronic or printed. Please also note that any views, opinions, 
> > conclusions or commitments expressed in this message are those of 
> > the individual sender and do not necessarily reflect the views of 
> > Fortinet, Inc., its affiliates, and emails are not binding on 
> > Fortinet and only a writing manually signed by Fortinet's General 
> > Counsel can be a binding commitment of Fortinet to Fortinet's 
> > customers or partners. Thank you. ***
> >   
> 
> How are using the Mellanox device with Windows Server? PCI passthrough or SR-IOV?

If using SR-IOV then you can't use the Mellanox device directly. You have to use the synthetic device. If you use vdev_netvsc pseudo-device, then it sets up a failsafe PMD, TAP and the VF (Mellanox) PMD for you.

The experimental way is to use the netvsc PMD which will manage VF if available.