Mon, Mar 04, 2019 at 05:59:04AM CET, pa...@mellanox.com wrote: > > >> -----Original Message----- >> From: netdev-ow...@vger.kernel.org <netdev-ow...@vger.kernel.org> On >> Behalf Of Jiri Pirko >> Sent: Wednesday, February 27, 2019 6:17 AM >> To: Jakub Kicinski <jakub.kicin...@netronome.com> >> Cc: da...@davemloft.net; oss-driv...@netronome.com; >> netdev@vger.kernel.org >> Subject: Re: [PATCH net-next 2/8] devlink: add PF and VF port flavours >> >> Tue, Feb 26, 2019 at 07:24:30PM CET, jakub.kicin...@netronome.com wrote: >> >Current port flavours cover simple switches and DSA. Add PF and VF >> >flavours to cover "switchdev" SR-IOV NICs. >> > >> >Example devlink user space output: >> > >> >$ devlink port >> >pci/0000:82:00.0/0: type eth netdev p4p1 flavour physical >> >pci/0000:82:00.0/10000: type eth netdev eth0 flavour pcie_pf pf 0 >> >pci/0000:82:00.0/10001: type eth netdev eth1 flavour pcie_vf pf 0 vf 0 >> >pci/0000:82:00.0/10002: type eth netdev eth2 flavour pcie_vf pf 0 vf 1 >> >A given port is of its parent device. >In current scenario, its PF or VF. >Hence it should be device attribute and not a port attribute.
I think that this works. You have VF_rep ports, PF_rep ports and PHYSICAL ports. In mlxsw for example, there are only PHYSICAL ports. In sr-iov world, there is also a PHYSICAL port on the eswitch. The others are either facing PF of VF. Looks accurate. I don't see any need for "devlink dev" flavour. >So devlink dev show command have to show what device flavour is. >Is it well known PCI VF or PF or something else. >It will show subdev device attribute and its parent PCI (PF/VF) devlink device. >So we should have device flovour as PCI_PF or PCI_VF or SUBDEV. > >Again VF number showcasing here is very restrictive model. >Every PF/VF/Subdev represents its own 'port' and it is connected to eswitch >'port'. >Instead of showing VF here, it must be this 'port' or 'link' number that gives >right view. >Which netdev represents which VF is already linked in the VF rep-netdev sysfs >property. I think you confuse the eswtich ports (in Jakub's output it's them) and the actual VF. > >So flavour should be something like 'hostport' and when port is registered for >the eswitch side it should be 'switchport'. >With this there is very clear picture of which hostport is connected to which >eswitch port. >Just like how we see in the physical world. >