Re: [dpdk-dev] [PATCH v1 0/3] Introduce new class for vDPA device drivers

[Thomas Monjalon <thomas@monjalon.net>]

10/01/2020 15:18, Xu, Rosen:
> From: Thomas Monjalon <thomas@monjalon.net>
> > 10/01/2020 03:38, Xu, Rosen:
> > > From: Matan Azrad <matan@mellanox.com>
> > > > From: Xu, Rosen <rosen.xu@intel.com>
> > > > > From: Thomas Monjalon <thomas@monjalon.net>
> > > > > > 09/01/2020 03:27, Xu, Rosen:
> > > > > > > From: Thomas Monjalon <thomas@monjalon.net>
> > > > > > > > 08/01/2020 13:39, Xu, Rosen:
> > > > > > > > > From: Matan Azrad <matan@mellanox.com>
> > > > > > > > > > From: Xu, Rosen
> > > > > > > > > > > Did you think about OVS DPDK?
> > > > > > > > > > > vDPA is a basic module for OVS, currently it will take
> > > > > > > > > > > some exception path packet processing for OVS, so it
> > > > > > > > > > > still needs to integrate
> > > > > > > > eth_dev.
> > > > > > > > > >
> > > > > > > > > > I don't understand your question.
> > > > > > > > > >
> > > > > > > > > > What do you mean by "integrate eth_dev"?
> > > > > > > > >
> > > > > > > > > My questions is in OVS DPDK scenario vDPA device
> > > > > > > > > implements eth_dev ops, so create a new class and move ifc
> > > > > > > > > code to this new class
> > > > > > is not ok.
> > > > > > > >
> > > > > > > > 1/ I don't understand the relation with OVS.
> > > > > > > >
> > > > > > > > 2/ no, vDPA device implements vDPA ops.
> > > > > > > > If it implements ethdev ops, it is an ethdev device.
> > > > > > > >
> > > > > > > > Please show an example of what you claim.
> > > > > > >
> > > > > > > Answers of 1 and 2.
> > > > > > >
> > > > > > > In OVS DPDK, each network device(such as NIC, vHost etc) of
> > > > > > > DPDK needs to be implemented as rte_eth_dev and provides
> > > > > > > eth_dev_ops
> > > > > such
> > > > > > > as
> > > > > > packet TX/RX for OVS.
> > > > > >
> > > > > > No, OVS is also using the vhost API for vhost port.
> > > > >
> > > > > Yes, vhost pmd is not a good example.
> > > > >
> > > > > > > Take vHost(Virtio back end) for example, OVS startups vHost
> > > > > > > interface like
> > > > > > this:
> > > > > > > ovs-vsctl add-port br0 vhost-user-1 -- set Interface
> > > > > > > vhost-user-1 type=dpdkvhostuser drivers/net/vhost implements
> > > > > > > vHost as
> > > > > rte_eth_dev
> > > > > > and integrated in OVS.
> > > > > > > OVS can send/receive packets to/from VM with
> > > > > > > rte_eth_tx_burst()
> > > > > > > rte_eth_rx_burst() which call eth_dev_ops implementation of
> > > > > > drivers/net/vhost.
> > > > > >
> > > > > > No, it is using rte_vhost_dequeue_burst() and
> > > > > > rte_vhost_enqueue_burst() which are not in ethdev.
> > > > > >
> > > > > > > vDPA is also Virtio back end and works like vHost, same as
> > > > > > > vHost, it will be implemented as rte_eth_dev and also be integrated
> > into OVS.
> > > > > >
> > > > > > No, vDPA is not "implemented as rte_eth_dev".
> > > > >
> > > > > Currently, vDPA isn't integrated with OVS.
> > > > >
> > > > > > > So, it's not ok to move ifc code from drivers/net.
> > > > > >
> > > > > > drivers/net/ifc has no ethdev implementation at all.
> > > > >
> > > > > For OVS hasn't integrated vDPA, it doesn't implement rte_eth_dev,
> > > > > but there are many discussions in OVS community about vDPA, some
> > > > > are from Mellanox, it seems vDPA port will be implemented as
> > > > > rte_eth_dev port in OVS in the near feature.
> > > > > https://eur03.safelinks.protection.outlook.com/?url=https%3A%2F%2F
> > > > > patc
> > > > > h
> > > > >
> > > >
> > work.ozlabs.org%2Fpatch%2F1178474%2F&amp;data=02%7C01%7Cmatan%
> > > > 4
> > > > >
> > > >
> > 0mellanox.com%7C9e84c2581e2f414e0aca08d794f22e8d%7Ca652971c7d2e4
> > > > d
> > > > >
> > > > 9ba6a4d149256f461b%7C0%7C0%7C637141640216181763&amp;sdata=TA%
> > > > 2F
> > > > >
> > 0zU495kXUqhC6eP09NDzBZfjJz1dbfkRcDpV%2BYAs%3D&amp;reserved=0
> > > > >
> > > > > Matan,
> > > > > Could you clarify how OVS integrates vDPA in Mellanox patch?
> > > > >
> > > > > >
> > > > > > Rosen, I'm sorry, these arguments look irrelevant, so I won't
> > > > > > consider them as blocking the integration of this patch.
> > > > >
> > > > > What I mentioned is not blocking the integration of this patch, I
> > > > > just want to get clarification from Matan how to integrate vDPA port in
> > OVS.
> > > >
> > > >
> > > > Hi
> > > >
> > > > OVS like any other application should use the current API of vDPA to
> > > > attach a probed vdpa device to a vhost device.
> > > > See example application /examples/vdpa.
> > > >
> > > > Here, we just introduce a new class to hold all the vDPA drivers, no
> > > > change in the API.
> > > >
> > > > As I understand, no vDPA device is currently integrated in OVS.
> > > >
> > > > I think it can be integrated only when a full offload will be
> > > > integrated since the vDPA device forward the traffic from the HW
> > > > directly to the virtio queue, once it will be there, I guess the
> > > > offload will be configured by the representor of the vdpa device(VF)
> > which is managed by an ethdev device.
> > > >
> > > >
> > > > Matan.
> > > >
> > > Hi,
> > >
> > > I'm still confused about your last sentence " the representor of the vdpa
> > device(VF) which is managed by an ethdev device".
> > > My understanding is that there are some connections and dependency
> > between rte_eth_dev and vdpa device?
> > > Am I right or any other explanations from you?
> > 
> > A vDPA port does not allow any ethdev operations (like rte_flow).
> > In order to configure some offloads on the device, OVS needs an ethdev port.
> > In Mellanox case, an ethdev VF representor port can be instantiated.
> > So we may have two ports for the same device:
> > 	- vDPA for data path with the VM
> > 	- ethdev for offloads control path
> 
> It's obviously that OVS needs these two functions of same device.
> In DPDK part, I have some concerns about how to scan, probe and bond these two ports
> of same device. Could you introduce it?
> As far as I know, each network device provided by DPDK is identified as a port in OVS,
> do you mind to take more clarification about how to connect or bond these two
> functions of same device in OVS part?

This is a different discussion.
You are asking a design review of vDPA integration in OVS.
I think it should be discussed separately.
And I have no time for such discussion currently. It will come later.

As it does not block this series, I will stop here, sorry.
Feel free to start a thread proposing how to integrate vDPA with offloads,
either in general, or specifically for IFC.