Host-side Interface Configuration
Multiple BlueField cards could connect to the same host machine. Each board has its own device directory (/dev/rshim<N>). The following are some guidelines on how to set up RShim networking properly in this case:
Each target should load only one backend (usb, pcie or pcie_lf)
The host RShim network interface should have different MAC and IP addresses, which can be configured with ifconfig as shown below, or saved in configuration:
$ ifconfig tmfifo_net0 192.168.100.1/24 hw ether 02:02:02:02:02:02
The Arm-side tmfifo interface should have unique MAC and IP addresses as well, which can be configured in the console
Multi-board Management Example
This example deals with two BlueField DPUs installed on the same server (the process is similar for more DPUs).
This example assumes that the RShim package has been installed on the host server.
Configuring Host Server Side
This example is relevant for CentOS/RHEL operating systems only.
Create a bf_tmfifo interface under /etc/sysconfig/network-scripts/. Run:
vim /etc/sysconfig/network-scripts/ifcfg-br_tmfifo
Inside ifcfg-br_tmfifo, insert the following content:
DEVICE="br_tmfifo" BOOTPROTO="static" IPADDR="192.168.100.1" NETMASK="255.255.255.0" ONBOOT="yes" TYPE="Bridge" NM_CONTROLLED="no"
Create a configuration file for the first BlueField DPU, tmfifo_net0. Run:
vim /etc/sysconfig/network-scripts/ifcfg-tmfifo_net0
Inside ifcfg-tmfifo_net0, insert the following content:
DEVICE=tmfifo_net0 BOOTPROTO=none ONBOOT=yes NM_CONTROLLED=no BRIDGE=br_tmfifo
Create a configuration file for the second BlueField DPU, tmfifo_net1. Run:
DEVICE=tmfifo_net1 BOOTPROTO=none ONBOOT=yes NM_CONTROLLED=no BRIDGE=br_tmfifo
Create the rules for the tmfifo interfaces. Run:
vim /etc/udev/rules.d/91-tmfifo_net.rules
Inside 91-tmfifo_net.rules, insert the following content:
SUBSYSTEM==
"net", ACTION=="add", ATTR{address}=="00:1a:ca:ff:ff:02", ATTR{type}=="1", NAME="tmfifo_net0", RUN+="/bin/sh -c '/usr/sbin/ifup tmfifo_net0 2>/dev/null || /sbin/ip link set dev tmfifo_net0 up'"SUBSYSTEM=="net", ACTION=="add", ATTR{address}=="00:1a:ca:ff:ff:04", ATTR{type}=="1", NAME="tmfifo_net1", RUN+="/bin/sh -c '/usr/sbin/ifup tmfifo_net1 2>/dev/null || /sbin/ip link set dev tmfifo_net1 up'"Restart the network for the changes to take effect. Run:
# /etc/init.d/network restart Restarting network (via systemctl): [ OK ]
Configuring BlueField DPU Side
BlueField DPUs arrive with the following factory default configurations for tmfifo_net0.
|
Address |
Value |
|
MAC |
00:1a:ca:ff:ff:01 |
|
IP |
192.168.100.2 |
Therefore, if you are working with more than one DPU, you must change the default MAC and IP addresses.
Updating MAC Address
This procedure is relevant for Ubuntu/Debian (sudo needed) and CentOS BFBs.
Open two console screen tabs per BlueField DPU. Run:
# sudo screen /dev/rshim<0|1>/console 115200
WarningScreen console RPM must be installed.
# sudo yum install screen
If the /dev/rshim device does not exist, the user should confirm they have RShim drivers installed. Run the following on the x86 host:
sudo dpkg -l rshim
Create a configuration file for tmfifo_net0 MAC address. Run:
# sudo vi /etc/bf.cfg
Inside bf.cfg, insert the new MAC:
NET_RSHIM_MAC=00:1a:ca:ff:ff:03
Apply the new MAC address. Run:
sudo bfcfg
Repeat this procedure for the second BlueField DPU (using a different MAC address).
NoteArm must be rebooted for this configuration to take effect. It is recommended to update the IP address before you do that to avoid unnecessary reboots.
Updating IP Address
For Ubuntu:
Access the file 50-cloud-init.yaml and modify the tmfifo_net0 IP address:
sudo vim /etc/netplan/50-cloud-init.yaml tmfifo_net0: addresses: - 192.168.100.2/30 ===>>> 192.168.100.3/30
Reboot the Arm. Run:
sudo reboot
Repeat this procedure for the second BlueField DPU (using a different IP address).
NoteArm must be rebooted for this configuration to take effect. It is recommended to update the MAC address before you do that to avoid unnecessary reboots.
For CentOS:
Access the file ifcfg-tmfifo_net0. Run:
# vim /etc/sysconfig/network-scripts/ifcfg-tmfifo_net0
Modify the value for IPADDR:
IPADDR=192.168.100.3
Reboot the Arm. Run:
reboot
Or perform netplan apply.
Repeat this procedure for the second BlueField DPU (using a different IP address).
NoteArm must be rebooted for this configuration to take effect. It is recommended to update the MAC address before you do that to avoid unnecessary reboots.
It is assumed that the commands in this section are executed with root (or sudo) permission.
The default MAC address is 00:1a:ca:ff:ff:01. It can be changed using ifconfig or by updating the UEFI variable as follows:
Log into Linux from the Arm console.
Run:
$ "ls /sys/firmware/efi/efivars".
If not mounted, run:
$ mount -t efivarfs none /sys/firmware/efi/efivars $ chattr -i /sys/firmware/efi/efivars/RshimMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c $ printf "\x07\x00\x00\x00\x00\x1a\xca\xff\xff\x03" > \ /sys/firmware/efi/efivars/RshimMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c
The printf command sets the MAC address to 00:1a:ca:ff:ff:03 (the last six bytes of the printf value). Either reboot the device or reload the tmfifo driver for the change to take effect.
The MAC address can also be updated from the server host side while the Arm-side Linux is running:
Enable the configuration. Run:
# echo "DISPLAY_LEVEL 1" > /dev/rshim0/misc
Display the current setting. Run:
# cat /dev/rshim0/misc DISPLAY_LEVEL 1 (0:basic, 1:advanced, 2:log) BOOT_MODE 1 (0:rshim, 1:emmc, 2:emmc-boot-swap) BOOT_TIMEOUT 300 (seconds) DROP_MODE 0 (0:normal, 1:drop) SW_RESET 0 (1: reset) DEV_NAME pcie-04:00.2 (ro) DEV_INFO BlueField-2(Rev 0) PEER_MAC 00:1a:ca:ff:ff:01 (rw) PXE_ID 0x00000000 (rw) VLAN_ID 0 0 (rw)
Modify the MAC address. Run:
$ echo "PEER_MAC xx:xx:xx:xx:xx:xx" > /dev/rshim0/misc
The BlueField-2 OOB interface is a gigabit Ethernet interface which provides TCP/IP network connectivity to the Arm cores. This interface is named oob_net0 and is intended to be used for management traffic (e.g. file transfer protocols, SSH, etc). The Linux driver that controls this interface is named mlxbf_gige.ko, and is automatically loaded upon boot. This interface can be configured and monitored by use of standard tools (e.g. ifconfig, ethtool, etc). The OOB interface is subject to the following design limitations:
Only supports 1Gb/s full-duplex setting
Only supports GMII access to external PHY device
Supports maximum packet size of 2KB (i.e. no support for jumbo frames)
The OOB interface can also be used for PXE boot. This OOB port is not a path for the BlueField-2 boot stream. Any attempt to push a BFB to this port will not work. Please refer to How to use the UEFI boot menu for more information about UEFI operations related to the OOB interface.
OOB Interface MAC Address
The MAC address to be used for the OOB port is burned into Arm-accessible UPVS EEPROM during the manufacturing process. This EEPROM device is different from the SPI Flash storage device used for the NIC firmware and associated NIC MACs/GUIDs. The value of the OOB MAC address is specific to each platform and is visible on the board-level sticker.
It is not recommended to reconfigure the MAC address from the MAC configured during manufacturing.
If there is a need to re-configure this MAC for any reason, follow these steps to configure a UEFI variable to hold new value for OOB MAC.:
The creation of an OOB MAC address UEFI variable will override the OOB MAC address defined in EEPROM, but the change can be reverted.
Log into Linux from the Arm console.
Issue the command ls /sys/firmware/efi/efivars to show whether efivarfs is mounted. If it is not mounted, run:
mount -t efivarfs none /sys/firmware/efi/efivars
Run:
chattr -i /sys/firmware/efi/efivars/OobMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c
Set the MAC address to 00:1a:ca:ff:ff:03 (the last six bytes of the printf value).
printf "\x07\x00\x00\x00\x00\x1a\xca\xff\xff\x03" > /sys/firmware/efi/efivars/OobMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c
Reboot the device for the change to take effect.
To revert this change and go back to using the MAC as programmed during manufacturing, follow these steps:
Log into UEFI from the Arm console, go to "Boot Manager" then "EFI Internal Shell".
Delete the OOB MAC UEFI variable. Run:
dmpstore -d OobMacAddr
Reboot the device by running "reset" from UEFI.
Log into Linux from the Arm console.
Issue the command ls /sys/firmware/efi/efivars to show whether efivarfs is mounted. If it is not mounted, run:
mount -t efivarfs none /sys/firmware/efi/efivars
Run:
chattr -i /sys/firmware/efi/efivars/OobMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c
Reconfigure the original MAC address burned by the manufacturer in the format aa\bb\cc\dd\ee\ff. Run:
printf "\x07\x00\x00\x00\x00\<original-MAC-address>" > /sys/firmware/efi/efivars/OobMacAddr-8be4df61-93ca-11d2-aa0d-00e098032b8c
Reboot the device for the change to take effect.
Supported ethtool Options for OOB Interface
The Linux driver for the OOB port supports the handling of some basic ethtool requests: get driver info, get/set ring parameters, get registers, and get statistics.
To use the ethtool options available, use the following format:
$ ethtool [<option>] <interface>
Where <option> may be:
<no-argument> – display interface link information
-i – display driver general information
-S – display driver statistics
-d – dump driver register set
-g – display driver ring information
-G – configure driver ring(s)
-k – display driver offload information
-a – q ueries the specified Ethernet device for pause parameter information
-r – re starts auto-negotiation on the specified Ethernet device if auto-negotiation is enabled
For example:
$ ethtool oob_net0
Settings for oob_net0:
Supported ports: [ TP ]
Supported link modes: 1000baseT/Full
Supported pause frame use: Symmetric
Supports auto-negotiation: Yes
Supported FEC modes: Not reported
Advertised link modes: 1000baseT/Full
Advertised pause frame use: Symmetric
Advertised auto-negotiation: Yes
Advertised FEC modes: Not reported
Link partner advertised link modes: 1000baseT/Full
Link partner advertised pause frame use: Symmetric
Link partner advertised auto-negotiation: Yes
Link partner advertised FEC modes: Not reported
Speed: 1000Mb/s
Duplex: Full
Port: Twisted Pair
PHYAD: 3
Transceiver: internal
Auto-negotiation: on
MDI-X: Unknown
Link detected: yes
$ ethtool -i oob_net0
driver: mlxbf_gige
version:
firmware-version:
expansion-rom-version:
bus-info: MLNXBF17:00
supports-statistics: yes
supports-test: no
supports-eeprom-access: no
supports-register-dump: yes
supports-priv-flags: no
# Display statistics specific to BlueField-2 design (i.e. statistics that are not shown in the output of "ifconfig oob0_net")
$ ethtool -S oob_net0
NIC statistics:
hw_access_errors: 0
tx_invalid_checksums: 0
tx_small_frames: 1
tx_index_errors: 0
sw_config_errors: 0
sw_access_errors: 0
rx_truncate_errors: 0
rx_mac_errors: 0
rx_din_dropped_pkts: 0
tx_fifo_full: 0
rx_filter_passed_pkts: 5549
rx_filter_discard_pkts: 4
IP Address Configuration for OOB Interface
The files that control IP interface configuration are specific to the Linux distribution. The udev rules file (/etc/udev/rules.d/92-oob_net.rules) that renames the OOB interface to "oob_net0" and is the same across all three major distributions (Yocto, CentOS, Ubuntu):
SUBSYSTEM=="net", ACTION=="add", DEVPATH=="/devices/platform/MLNXBF17:00/net/eth[0-9]", NAME="oob_net0"
The files that control IP interface configuration are slightly different across the two major distributions (CentOS, Ubuntu):
CentOS configuration of IP interface:
Configuration file for "oob_net0" interface: /etc/sysconfig/network-scripts/ifcfg-oob_net0
For example, use the following to enable DHCP:
NAME="oob_net0" DEVICE="oob_net0" NM_CONTROLLED="yes" PEERDNS="yes" ONBOOT="yes" BOOTPROTO="dhcp" TYPE=Ethernet
For example, to configure static IP use the following:
NAME="oob_net0" DEVICE="oob_net0" IPV6INIT="no" NM_CONTROLLED="no" PEERDNS="yes" ONBOOT="yes" BOOTPROTO="static" IPADDR="192.168.200.2" PREFIX=30 GATEWAY="192.168.200.1" DNS1="192.168.200.1" TYPE=Ethernet
For Ubuntu configuration of IP interface, please refer to section "Network Interface Configuration".