Ipmid not work on eth1 when enable NCSI.

[Thu Nguyen]

Hi All,


I'm verifying NCSI with my Mt.Jade platform.

In Mt.Jade platform, we have the shared NIC in eth0 interface and 
dedicated RGMII in eth1 interface.

I used below steps to enable NC-SI in Mt.Jade.

1. Add definition in DTS for MAC0:

&mac0 {
       status = "disabled";
       status = "okay";
       pinctrl-names = "default";
       pinctrl-0 = <&pinctrl_rmii1_default>;
       clocks = <&syscon ASPEED_CLK_GATE_MAC1CLK>,
                <&syscon ASPEED_CLK_MAC1RCLK>;
       clock-names = "MACCLK", "RCLK";
       use-ncsi;
};

2. Add build configure to enable OEM Command to get MAC for NCSI.

CONFIG_NCSI_OEM_CMD_GET_MAC=y

After these steps, My system can detect OCP card, can get MACs and both 
eth0 and eth1 interfaces are up.

I can access BMC Web use both ip of eth0 or eth1. The same for Redfish.

BUT I can only use ipmitool with eth0. Ipmitool command not work with 
eth1 interface.


Check the setting of phosphor-ipmid-net in

meta-phosphor/recipes-phosphor/ipmi/phosphor-ipmi-config/channel_access.json

meta-phosphor/recipes-phosphor/ipmi/phosphor-ipmi-config/channel_config.json

I can see that ipmid is configured to listen on both interface eth0 and 
eth1.

Do I miss any configuration here?


Check the journal log, I saw some errors and warning from NCSI:

Dec 23 10:14:50 mtjade kernel: ftgmac100 1e660000.ethernet eth0: NCSI: 
Handler for packet type 0x82 returned -19

Dec 23 10:15:05 mtjade kernel: ftgmac100 1e660000.ethernet eth0: NCSI: 
'bad' packet ignored for type 0x8b
Dec 23 10:15:05 mtjade kernel: ftgmac100 1e660000.ethernet eth0: NCSI: 
'bad' packet ignored for type 0x8b
Dec 23 10:15:05 mtjade kernel: ftgmac100 1e660000.ethernet eth0: NCSI: 
'bad' packet ignored for type 0x8b
Dec 23 10:15:05 mtjade kernel: ftgmac100 1e660000.ethernet eth0: NCSI: 
'bad' packet ignored for type 0x8b

Because NC-SI interface is up, so I just ignore these errors.


Regards.

Thu Nguyen.