Brian, Are you enabling pause frames? It is required when transmitting to the USRP at high rates with 100 GbE.
https://files.ettus.com/manual/page_transport.html#transport_udp_linux Unfortunately, as far as I know, it's not yet enabled by default for some technical reasons. So be sure to turn it on after each FPGA reload or reboot. You should be able to read those registers by doing raw memory read/writes to the system bus. The register offsets are here: https://github.com/EttusResearch/uhd/blob/master/fpga/usrp3/lib/rfnoc/xport_sv/eth_regs.vh The base address for QSFP0 I think is 0x12_0000_8000, so REG_CHDR_DROPPED, for example, should be at 0x12_0000_9030. The base address for QSFP1 should be 0x12_0001_8000. For example, running this should return the bottom part of the MAC address: devmem2 0x1200009010 Hopefully I've got the details right. I'm traveling at the moment, so I can't easily verify. Wade On Thu, Aug 14, 2025 at 7:04 PM Brian Padalino <bpadal...@gmail.com> wrote: > On Thu, Aug 14, 2025 at 7:38 PM Nikos Balkanas <nbalka...@gmail.com> > wrote: > >> Yeah, >> >> It makes sense. But just to be on the safe side >> "I could not ping the interface (Destination unreachable) and packets >> stopped flowing through the interface from the host machine" >> wouldn't at least give it a try? Seems you have lost your routing tables >> to get Destination Unreachable over a static route! >> When you get the problem what does the arp command show? >> > > Before the issue, arp shows the connection appropriately and pings work > fine. After flooding, arp shows (incomplete) and the ping responds with a > Destination Host Unreachable message. > > With a static arp entry, the packet keep flowing (better!) but the pings > still get lost. Internally I see tready fall at the output of the ethernet > transport adapter into the crossbar and never come back. The CHDR dropped > count continues in the eth_ipv4_interface. > > I think going with the static entry just masks the issue that things are > getting gummed up somewhere. I think I am just surprised that the ICMP > stuff stops working since I thought the CHDR and CPU paths diverge before > the crossbar so things like pings should still work even if the CHDR > pipeline is stalled. > > Brian > >> _______________________________________________ > USRP-users mailing list -- usrp-users@lists.ettus.com > To unsubscribe send an email to usrp-users-le...@lists.ettus.com >
_______________________________________________ USRP-users mailing list -- usrp-users@lists.ettus.com To unsubscribe send an email to usrp-users-le...@lists.ettus.com
