[enhacement] Disable Intel firmware embedded LLDP agent before collecting LLDP data

           *-network:0
                description: Ethernet interface
                product: Ethernet Controller X710 for 10GbE SFP+
                vendor: Intel Corporation
                physical id: 0
                bus info: pci@0000:01:00.0
                logical name: eno1
                version: 01
                serial: 24:6e:96:XX:XX:XX
                size: 10Gbit/s
                width: 64 bits
                clock: 33MHz
                capabilities: pm msi msix pciexpress vpd bus_master cap_list rom ethernet physical fibre
                configuration: autonegotiation=off broadcast=yes driver=i40e driverversion=1.4.25-k duplex=full firmware=6.00 0x800034e6 18.3.6 latency=0 link=yes multicast=yes slave=yes speed=10Gbit/s

Hi Nobuto,

What's the best way to discover this interface automatically? Is there a file we can check somewhere under sys/kernel to allow us to identify this interface?

> What's the best way to discover this interface automatically? Is there a file we can check somewhere under sys/kernel to allow us to identify this interface?

The path is /sys/kernel/debug/i40e/<PCI_ID>/command like:
/sys/kernel/debug/i40e/0000:05:00.0/command

I suppose MAAS can write "lldp stop" whenever the paths exist. Because it's not persistent across reboots so it safely can be modified in the commissioning phase.

This bug has not seen any activity in the last 6 months, so it is being automatically closed.

If you are still experiencing this issue, please feel free to re-open.

MAAS Team

I'm developing a tool that will rely on lldp data collected by maas to draw a network map to catch errors and miscabling during HAP phase of cloud install in customers. It is very important that this gets fixed.

I know I can upload scripts to fix that (have it working right now), but this severely compromises the workflow because we normally use the "fce" tool (cpe-foundation) that does the maas installation and configuration for us and it would require us to:

1) patch fce for a workaroung (idea that the FCE team does not like, unless we clearly declare this is a bug that is going to get fixed in the future)

2) make the user go out of automatic process and inject an extra script in maas in the middle of the installation.

My goal is to have this fixed upstream as soon as possible and to get fce to workaround it until new version is generally available.

The fix seems simple enough. I think the best place to fix it is to send commands to the card on the same "lldp install" script, right before enabling the deamon.

I understand that we probably lack the hardware in our lab to test this, but I have access to clients hardware and have tested this, it works. And it is simple enough to have no consequences for any other case.

Shoud I just go ahead and submit an MP?

Just FYI, the MP was submitted, still pending.

https://code.launchpad.net/~andre-ruiz/maas/+git/maas-fix-intel-cna-lldp/+merge/389824

Technically the 'standard' way to enable/disable i40e fw lldp is to use ethtool, e.g.:

root@dwalin:~# ethtool --show-priv-flags enp4s0f0 | grep fw-lldp
disable-fw-lldp : off

root@dwalin:~# ethtool --set-priv-flags enp4s0f0 disable-fw-lldp on

root@dwalin:~# ethtool --show-priv-flags enp4s0f0 | grep fw-lldp
disable-fw-lldp : on

The debugfs api to i40e is probably not going to change though, it should be ok to use.

Thanks Dan, I did not know about that way. I agree it would be more interesting, and I could update the patch if needed, but since the debug interface is stable I'll leave it like that for now. We can revisit this in the future if needed.