Using Tioga Pass Channel onboard shared-NIC causes BMC reset

[MennoCB]

Hi,

I am using the Tioga Pass Channel version from Wiwynn and I am trying to use the on-board Intel I210 with RJ45 connector for management, instead of using NCSI through the Mezzanine card (Mezzanine works just fine).
It seems to detect the NIC but when I try to configure an IP on it with ifconfig or through interfaces, the BMC just resets/reboots without any message the console.

[ 18.675494] ftgmac100 ftgmac100.0 eth0: irq 2, mapped at dca40000
[ 18.687826] ftgmac100 ftgmac100.0 eth0: generated random MAC address b6:8b:68:aa:f0:30
[ 18.703743] dev_attr_powerup_prep_host_id registered
[ 18.713748] ncsi_nl_socket_init: created netlink socket
[ 18.724270] ftgmac: aen work queue created for eth0
[ 18.734789] ftgmac100 ftgmac100.1 eth1: irq 3, mapped at dca80000
[ 18.747487] ftgmac100 ftgmac100.1 eth1: generated random MAC address 66:61:4e:a1:8a:9e
[ 18.763413] dev_attr_powerup_prep_host_id registered
[ 18.773425] ftgmac: aen work queue created for eth1

Configuring network interfaces... [ 26.995441] Found NCSI NW Controller at (0, 0)
[ 27.506454] NCSI: Mezz Vendor = Mellanox
[ 28.015603] NCSI: MAC 24:8A:07:3F:96:B0:
[ 28.025810] Using NCSI Network Controller (0, 0)
[ 28.400902] random: nonblocking pool is initialized

This option is described in the Facebook 2S Server Tioga Pass specification, secion 9.1:
Option 3: Shared-NIC uses RMII/NCSI interfaces to pass management traffic on data network of Intel® I210-AT. Intel® I210-AT has 10/100/1000 MDI interface to RJ45.

OpenBMC Release fbtp-25474cc

root@bmc-oob:~# fruid-util all

FRU Information : Mother Board
--------------- : ------------------
Chassis Type : Rack Mount Chassis
Chassis Part Number :
Chassis Serial Number :
Chassis Custom Data 1 : CPU:
Chassis Custom Data 2 : CPU:
Board Mfg Date : Wed Jun 6 03:55:00 2018
Board Mfg : Wistron
Board Product : Tioga-Pass Channel
Board Serial : B5501N01000182200010J0A1
Board Part Number : B91.01N10.0001
Board FRU ID : N/A
Product Manufacturer : Wiwynn
Product Name : Tioga-Pass Channel
Product Part Number :
Product Version : FAB1,SA
Product Serial :
Product Asset Tag :
Product FRU ID : N/A
Product Custom Data 1 :
Product Custom Data 2 : PVT
Product Custom Data 3 : 2016-03-24T00:00:00
Failed print FRUID for NIC Mezzanine
Check syslog for errors!
Failed print FRUID for FRU content on the riser slot 2
Check syslog for errors!
Failed print FRUID for FRU content on the riser slot 3
Check syslog for errors!
Failed print FRUID for FRU content on the riser slot 4
Check syslog for errors!

Thanks

[MennoCB]

Also after installing the OS, sometimes when I reboot the OS this message appears on the BMC:
[ 2214.445881] ftgmac100 ftgmac100.0 eth0: rx crc err
[ 2214.455650] ftgmac100 ftgmac100.0 eth0: rx runt

And then BMC is not reachable on NCSI, no reply on ping, need to do "mc reset cold" from the OS to make it reboot.

[MennoCB]

Just build another fbtp and now it's working

[ 18.888541] ftgmac100 ftgmac100.1 eth1: irq 3, mapped at dca80000
[ 18.901234] ftgmac100 ftgmac100.1 eth1: generated random MAC address e6:71:91:3b:8f:4b
[ 18.927180] ftgmac: aen work queue created for eth1

Still doesn't use it though, when I set metric in /etc/network/interfaces for eth0 and eth1 it's not taking that to the routing table... but it's one step closer.

The 10G Mezz is eth0 and on-board is 1G eth1, so it's still not using the 1G RJ45 interface ;-)