Hope summary below could answer your questions.

From: lukego at gmail.com [mailto:luk...@gmail.com] On Behalf Of Luke Gorrie
Sent: Monday, November 24, 2014 7:58 PM
To: Thomas Monjalon
Cc: Zhou, Danny; dev at dpdk.org; Fastabend, John R
Subject: Re: [dpdk-dev] bifurcated driver

On 5 November 2014 at 14:00, Thomas Monjalon <thomas.monjalon at 
6wind.com<mailto:thomas.monjalon at 6wind.com>> wrote:
It seems to be close to the bifurcated driver needs.
Not sure if it can solve the security issues if there is no dedicated MMU
in the NIC.

I feel we should sum up pros and cons of
        - igb_uio
        - uio_pci_generic
        - VFIO
        - ibverbs
        - bifurcated driver

I am also curious about the pros and cons of the bifurcated driver compared 
with SR-IOV.
DZ: I have a slide to compare all of them except for ibverbs, but system does 
not allows me to paste a captured picture. I can send it to you in a separated 
email.

What are the practical differences between running a bifurcated driver vs. 
running SR-IOV mode where the kernel owns the PF and userspace applications own 
the VFs?
DZ: SRIOV can be treated as a rx/tx queue partition approach following PCIe 
spec, but it diffs from bifurcated driver from as illustrated below:

a)      On ixgbe, each VF can only have at maximum two rx/tx qpairs, while 
bifurcated driver supports allocating multiple rx/tx qpairs per user space 
request.

b)      Each PF/VF has a dedicated MAC address and/or VLAN ID, and L2 switch 
inside a NIC distributes packets to PF/VF based on MAC address or VLAN ID. 
While bifurcated driver

builds on top of in_NIC flow director which allows up-to 8K L3/L4 filters (e.g. 
5-tuple to rx queue), rather than fixed L2 filter, to distribute traffics to 
either kernel or user space.

c)       VF does not have dedicated flow director.

Specifically, could I run the ixgbe driver in the kernel (max_vfs=N), control 
it via ethtool, and then access the queues via userspace VF drivers? If so, how 
would this differ from the bifurcated driver?
DZ: Yes you can do this. But L3/L4 filters you setup to PF (VF does not support 
it) via ethtool to flow director takes no effect if the packet?s DMAC matches a 
VF? MAC, as the packet will be distributed by L2 switch to the VF that rx queue 
belongs to. Ixgbe does not support per-VF flow director that allows you setup 
filters distributing packet to two rx queue inside a VF.

Cheers,
-Luke


Reply via email to