That's strange; one interface can have multiple ports associated,
but each port should only be attached to one interface. The
default design should connect one interface to each SFP+ port,
with nothing connected to the RJ45 port (that port goes directly
to the zynq PL). I can test on my ZCU106 to see if I can
replicate what you're seeing, although I might need a few more
details on your setup.
Alex Forencich
--
You received this message because you are subscribed to the Google Groups "corundum-nic" group.
To unsubscribe from this group and stop receiving emails from it, send an email to corundum-nic...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/corundum-nic/061df180-d67d-4bde-966b-47c2d2763dd7n%40googlegroups.com.
My suggestion is to buy a 10G SFP+ PCIe NIC and an SFP+ DAC.
Currently, the SFP+ modules only run at 10 Gbps and don't
(currently) support negotiating down, so you would need 10GBASE-T
adapters and a 10G BASE-T NIC. However, that is probably far more
expensive than picking up a used 10G SFP+ NIC and DAC off of
ebay. Although it depends on what your final application would
be. If you ultimately want to run at 1 Gbps, then it is possible
to swap out the 10GBASE-R PHYs for 1000BASE-X PHYs on the FPGA,
and then you should be able to use gigabit RJ45 to SFP adapters.
I have not tested the corundum datapath at 8 bits, I may need to
tweak some parameters to get it to work properly. But you can
just use AXI stream width converters for the time being.
Also, bear in mind that linux tries to be smart and will route
the packets internally if it can. Most likely none of your test
packets left the host. There are a few solutions for this. Most
reliable is to use two machines. Another option is to use network
namespaces. I have a script for that:
https://github.com/alexforencich/verilog-ethernet/blob/master/scripts/dev-netns-shell.sh
. You run that script with the interface name as an argument, and
then it will create a namespace, add the interface to the
namespace, and then drop you into a shell running in that
namespace. It's also smart enough to manage the namespace nicely
so you can run multiple instances of the script attached to the
same interface. From there, you cannot see any of the other NICs,
so traffic is forced to use the associated NIC instead of getting
routed internally. So you can try running your ssh client from
there as a test case. My guess is that if you repeat the
experiment, you won't be able to connect at all because the
gigabit port on the ZCU106 is not used.
Alex Forencich
To view this discussion on the web visit https://groups.google.com/d/msgid/corundum-nic/b4ea05e7-ad54-497b-8ede-d78b86ef7d81n%40googlegroups.com.