TDA4VM-Q1: TDA4 connects to Switch via PCIE

Hi Yanggc,

By "Marvel's switch", I assume it is a PCIe switch to increase the number of PCIe slots, instead of a PCIe network switch to increase number of ethernet ports.

If it is a PCIe switch, then TDA4VM in particular has a silicon errata when using a PCIe switch. It can be worked around using method 3 in this FAQ: https://e2e.ti.com/support/processors-group/processors/f/processors-forum/1264686/faq-j721exsomxevm-how-to-work-around-errata-i2086-for-pcie-switches-and-multifunction-devices-for-j721e-tda4vm-dra829

Regards,

Takuma

Hi Yanggc,

I may have had a misunderstanding. It sounds like you are using a network switch to increase number of network ports instead of a PCIe switch to increase PCIe ports.

If it is a network switch connected through PCIe, then:

1. Hardware-wise, nothing special and no different than connecting any PCIe card, unless the particular network card you have has special out-of-bound signals like an extra enable pin (very rare, and should be mentioned in your network card's data sheet if there are such signals).
2. Software-wise:
   1. PCIe controller needs to be defined in device tree.
      1. If using the default TI EVM device tree with default EVM board, then PCIe0, 1, and 2 controllers are already defined and no extra work needed: https://git.ti.com/cgit/ti-linux-kernel/ti-linux-kernel/tree/arch/arm64/boot/dts/ti/k3-j721e-common-proc-board.dts?h=ti-linux-6.6.y#n931 . This will be enough for the network switch to get enumerated under PCIe bus and show up under tools like lspci.
      2. For custom boards, you will need to define the PCIe instance you are pinning out, and change reset gpio pin to whichever pin you decide to use in your custom board design.
   2. In addition to device tree, you may need to add in extra kernel configs for your particular network switch.
      1. You will need to find the kernel config required for your device by searching online or contacting the designer of that network switch.
      2. Then, you can follow this documentation to configure the kernel: https://software-dl.ti.com/jacinto7/esd/processor-sdk-linux-jacinto7/10_01_00_04/exports/docs/linux/Foundational_Components_Kernel_Users_Guide.html#configuring-the-kernel 
      3. Once kernel config includes the driver for your particular network switch, and the new kernel is flashed, this will automatically bind to the PCIe device of the network switch. You may use "lspci -k" to check if the automatically binding occurred. Below is an example of a network card bound to Intel's iwlwifi kernel driver enumerated under lspci, and you should see something similar for your network switch:
         
      4. Once kernel driver is bound, this will allow the network switch to get enumerated under ifconfig or ip and allow you to use network tools such as iperf, ping, and any other network tools for testing.

Regards,

Takuma

Hi，

Hi Yanggc,

Understood, and best of luck!

Regards,

Takuma