[dpdk-dev] Question on documentation / Mellanox ConnectX-3

[Marcelo Ricardo Leitner]

On Mon, Apr 09, 2018 at 05:07:51PM +0200, Adrien Mazarguil wrote:
> On Sun, Apr 08, 2018 at 06:27:50PM -0300, Marcelo Ricardo Leitner wrote:
> > On Thu, Apr 05, 2018 at 11:26:21AM +0200, Adrien Mazarguil wrote:
> > > On Tue, Apr 03, 2018 at 02:59:38PM -0300, Marcelo Ricardo Leitner wrote:
> > > > Hi,
> > > >
> > > > http://docs.openvswitch.org/en/latest/howto/dpdk/ says:
> > > >
> > > > Some NICs (i.e. Mellanox ConnectX-3) have only one PCI address
> > > > associated with multiple ports. Using a PCI device like above won’t
> > > > work. Instead, below usage is suggested:
> > > >
> > > > $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
> > > >     options:dpdk-devargs="class=eth,mac=00:11:22:33:44:55:01"
> > > > $ ovs-vsctl add-port br0 dpdk-p1 -- set Interface dpdk-p1 type=dpdk \
> > > >     options:dpdk-devargs="class=eth,mac=00:11:22:33:44:55:02"
> > > >
> > > > But these MACs are 7 bytes long. Seems the idea was to mention the two
> > > > incremental MAC addresses that the ports have, and thus the ':55'
> > > > should have been removed from there, right?
> > > >
> > > > Reading the code, it doesn't seem prepared to handle the extra byte in
> > > > any (special) way.
> > >
> > > After a quick glance at the original patch [1], I confirm it looks like a
> > > mistake in the OVS documentation. MAC addresses should be 6 bytes long, the
> > > 7th byte is not a workaround to identify a physical port like I initially
> > > thought.
> > >
> > > As you pointed out, since default MAC addresses on mlx4 ports are normally
> > > incremental, documentation should read something like:
> > >
> > >  $ ovs-vsctl add-port br0 dpdk-p0 -- set Interface dpdk-p0 type=dpdk \
> > >      options:dpdk-devargs="class=eth,mac=00:11:22:33:44:55"
> > >  $ ovs-vsctl add-port br0 dpdk-p1 -- set Interface dpdk-p1 type=dpdk \
> > >      options:dpdk-devargs="class=eth,mac=00:11:22:33:44:56"
> > >
> > > [1] https://github.com/openvswitch/ovs/commit/5e7588186839
> >
> > Thanks Adrien. Question now is, who can fix the doc?
>
> Someone should submit a request or patch pointing to this thread to the
> documentation maintainer of OVS. Since it's a different project there's not
> much I can do from the DPDK side.

Makes sense. Thanks.

  Marcelo