Known Issues

Workaround: Create or append the following lines to the existing bf.cfg:

            

            
BMC_PATH=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/bmc
BMC_IMAGE=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/bmc/bf3-bmc-fw.fwpkg
CEC_PATH=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/cec
CEC_IMAGE=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/cec/bf3-cec-fw.fwpkg
DPU_GI_PATH=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/bmc
DPU_GOLDEN_IMAGE=/root/bfb/upgrade/initramfs/usr/lib/firmware/mellanox/bmc/bf3-bmc-preboot-install.bfb
        

Use this bf.cfg file with bfb-install/doca-installer or OobUpdate.

Workaround:

1. Set Boot option enable attribute to false.

2. Power reset DPU.

3. Do not reboot or power reset DPU for a second time. If reboot/power reset DPU does occur, the boot option enable attribute will be recovered to true on UEFI generated Boot Options, but this issue does not affect legacy boot options.

4. Completely disable boot options getting reset by UEFI by using `SYS_DISABLE_AUTO_BOOT_REFRESH=TRUE` configuration in bf.cfg.

Workaround:

1. Power-cycle BlueField.
2. Repeat push BFB.

Workaround: Attempt the following:

1. To update DOCA only, disable ALL repos:

   Copy

   Copied!

               
   
               
   dnf --disablerepo='*' --enablerepo='kubernetes,doca' -y update
           

2. To update OS components, exclude rdma-core using one of the following options:

   • Option 1 – Disable rdma-corem from dnf.conf using excludepkgs=rdma-core*oe2203sp3*
   • Option 2 – Disable rdma-core from openEuler.repo file under OS and everything using exclude=rdma-core*
   • Option 3 – yum update -x rdma-core

Workaround: Increase the CMA limit or cancel its use using the kernel's CMD line parameters:

• Add the parameter cma=256M to increase the CMA limit to 256MB; or
• Add the parameter cma=0 to disable the use of CMA

Workaround: Use the following command instead:

            

            
echo 0 > /sys/kernel/mm/hugepages/hugepages-<Size>/nr_hugepages
        

Description: On CentOS 7.6 with kernel 4.19, bringing up OVS bridge interface causes call traces:

            

            
WARNING: CPU: 5 PID: 14339 at kernel/rcu/tree_plugin.h:342 rcu_note_context_switch+0x48/0x538
        

Workaround: Set PCIE_HAS_VFIO=0 and PCIE_HAS_UIO=0 in /etc/rshim.conf and restart the RShim driver. Then re-run the mlxfwreset command.

If host Linux kernel lockdown is enabled, then manually unbind the RShim driver before mlxfwreset and bind it back after mlxfwreset:

            

            
echo "DROP_MODE 1" > /dev/rshim0/misc
mlxfwreset <arguments>
echo "DROP_MODE 0" > /dev/rshim0/misc
        

Description: UEFI synchronous exception is observed at address 0x479B7xxxx where the UEFI module names are not printed. See the following example:

            

            
ERR[UEFI]: PC=0x479B78480(B4000040 3900001F A94153F3 F94013F5)
ERR[UEFI]: PC=0x479B78480
ERR[UEFI]: PC=0x479B7E684
ERR[UEFI]: PC=0x47A0E93F4
ERR[UEFI]: PC=0x47A0E9608
        

Description: On Debian 12, the first boot after BFB installation may fail with the following kernel panic:

            

            
[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000100 ]
        

Workaround: Reset the DPU using the RShim interface:

            

            
echo "SW_RESET 1" > /dev/rshim0/misc
        

Workaround: The following are possible options –

• Boot the BFB file doca_2.5.0_bsp_4.5.0_ubuntu_22.04-1.23-10.prod.bfb to update the platform software
• Download, compile, and install latest mlxbf-bootctl command from GitHub

• Edit default.bfb by using the mlx-mkbfb command to incorporate the platform-specific images and filtering out unused images. Example for a BlueField-2 device:

  Copy

  Copied!

              
  
              
  $ mlx-mkbfb -x default.bfb
  $ mlx-mkbfb \
  --bl2r-v1=dump-bl2r-v1 \
  --bl2r-cert-v1=dump-bl2r-cert-v1 \
  --bl2-v1=dump-bl2-v1 \
  --bl2-cert-v1=dump-bl2-cert-v1 \
  --bl31-v1=dump-bl31-v1 \
  --bl31-cert-v1=dump-bl31-cert-v1 \
  --bl31-key-cert-v1=dump-bl31-key-cert-v1 \
  --bl33-v0=dump-bl33-v0 \
  --bl33-cert-v1=dump-bl33-cert-v1 \
  --bl33-key-cert-v1=dump-bl33-key-cert-v1 \
  --boot-acpi-v0=dump-boot-acpi-v0 \
  --boot-args-v0=dump-boot-args-v0 \
  --boot-desc-v0=dump-boot-desc-v0 \
  --boot-path-v0=dump-boot-path-v0 \
  --ddr_ini-v1=dump-ddr_ini-v1 \
  --ddr-cert-v1=dump-ddr-cert-v1 \
  --ddr_ate_imem-v1=dump-ddr_ate_imem-v1 \
  --ddr_ate_dmem-v1=dump-ddr_ate_dmem-v1 \
  --snps_images-v1=dump-snps_images-v1 \
  --trusted-key-cert-v1=dump-trusted-key-cert-v1 \
  default_min.bfb
          

Description: Symmetric pause must be enabled in the DHCP server for the OOB to be able to reliably get an IP address assigned.

Description: On Debian 12, Arm ports remain in Legacy mode after multiple Arm reboot iterations. The following error message appears in /var/log/syslog:

            

            
mlnx_bf_configure[2601]: ERR: Failed to configure switchdev mode for 0000:03:00.0 after 61 retries
        

Workaround: Run:

            

            
$ echo SET_MODE_RETRY_NUM=300 >> /etc/mellanox/mlnx-bf.conf
$ reboot
        

Workaround: Create /etc/bf.cfg with the following lines, then run bfcfg to recreate the PXE boot entries:

            

            
BOOT0=DISK
BOOT1=NET-NIC_P0-IPV4
BOOT2=NET-NIC_P0-IPV6
BOOT3=NET-NIC_P1-IPV4
BOOT4=NET-NIC_P1-IPV6
BOOT5=NET-OOB-IPV4
BOOT6=NET-OOB-IPV6
        

Workaround: Use netplan configuration. For example

            

            
# cat /etc/netplan/tmfifo_net0.yaml
network:
  version: 2
  renderer: networkd
  ethernets:
    tmfifo_net0:
      addresses:
      - 192.168.100.1/30
      dhcp4: false
        

Then run "netplan apply".

Description: After rebooting a BlueField-3 DPU running Rocky Linux 8.6 BFB, the kernel log shows the following error:

            

            
[    3.787135] mlxbf_gige MLNXBF17:00: Error getting PHY irq. Use polling instead
        

This message indicates that the Ethernet driver will function normally in all aspects, except that PHY polling is enabled.

Description: When trying to perform a PXE installation when UEFI Secure Boot is enabled, the following error messages may be observed:

            

            
error: shim_lock protocol not found.
error: you need to load the kernel first.
        

Workaround: Use ib_core module parameter netns_mode=0. For example:

            

            
echo "options ib_core netns_mode=0" >> /etc/modprobe.d/mlnx-bf.conf
        

Workaround: Set OPENSSL_CONF=/etc/ssl/openssl.cnf.orig before using the curl utility.

For example:

            

            
# OPENSSL_CONF=/etc/ssl/openssl.cnf.orig curl -O https://tpo.pe/pathogen.vim
        

Description: The host may crash when the number of PCIe devices overflows the PCIe device address. According to the PCIe spec, the device address space is 8 bits in total—device (5 bits) and function (3 bits)—which means that the total number of devices cannot be more than 256.

The second PF maximum number of VFs is limited by the total number of additional PCIe devices that precedes it. By default, the preceding PCIe devices are 2 PFs + RShim DMA + 127 VFs of the first PF. This means that the maximum valid number of VFs for the second port will be 126.