On our new custom carrier board I have connected the RMII port to a Micrel/Microchip KSZ8863RLL 3 port ethernet switch - Using the 2nd Ethernet documentations as reference.
The VF50 is configured for DHCP, but ends up with an internally assigned IP address as it cannot get a network connection through the KSZ8863.
From other posts it seems that others have had issues with the MDC/MDIO interface that allows management of the device at the other end of the RMII interface - When looking at the MDC and MDIO lines with a logic analyzer I can see the MDC beeing active for aprox. 1 second (780 kHz) but with no data on the MDIO apart from one slow fall and rise that looks like noise to me.
I have loaded the registry changes for activating the 2nd RMII port and the second ethernet icon shows up in the WEC 6 GUI.
I have checked the GPIO settings with Vybrid GPIO V1.2 and found that the RMII interface has indeed been activated with MDC on GPIO 71 and MDIO on GPIO 98 (BL_ON has been moved to GPIO 106).
The KSZ8863RLL chip is working on the 2 other ports allowing a connection through the switch.
Any suggestions on how to investigate from here would be much appreciated.
Can you please provide (uploading on share.toradex.com if needed) the schematics of the 2nd ethernet connection to the module and the registry settings you added to enable the 2nd controller?
For the registry changes I have used ‘2nd_ethernet_regisrty.reg’ file from the Toradex website.
I have made 2 changes as the default IP address and gateway was entered with spaces (0 . 0 . 0 . 0) which I have changed to 0.0.0.0.
The Bind parameter was entered with spaces to which I have also changed.
I can’t see the correspondence between the signals you routed to the KSZ8863 and the SO-DIMM pins, that would help to understand the problem.
I checked the registry file on our developer website:
And .reg sample does not contain spaces in the IPs (they are configured as hex strings and have 0 and ‘.’ characters.
Can you please share also your registry entries?
The RMII interface signals from the KSZ8863RLL to the VF50 are shown on page 1 to the far left except for RMII0_CRS_DV (101) and RMII0_RxD1 (103) that are in the top of the schematics.
I see that you have DHCP enabled for the 2nd ENET instance, is that correct? Are both instances connecting to the same network?
Can you please use the GPIO tool to check how the relevant pins are configured on your device?
Can you try to use wireshark to figure out if the non-working ethernet is sending out packets? You may connect that directly to your PC using a crossed cable and monitor all packets.
I have build 2 identical boards and tested with 2 different VF50 modules and the result is the same.
I have used Wireshark to see if a DHCP/ARP request was being send from the system but that is not the case.
I have not yet tried to connect a logic analyzer directly to internal MDIO lines on the Iris Vf50 but I would assume that both ethernet interfaces on the NXP Freescale CPU would at least read the 2 manufacture ID / chip ID memory locations via the MDIO bus.
I have checked the using the GPIO Tool that the 2nd ethernet port has indeed configured the corresponding I/Os for RMII and the MDIO interface - Is there a way to configure the address used on the MDIO bus ?
If I’m not mistaken there was an issue in the Linux image where the 2nd ethernet port used the MDIO lines of the 1st ethernet port by mistake, but may never have been an issue under WEC 6.
I have tested 2 Ethernet interfaces on Vybrid, WinCE6. It is working with the additional registry. Could you please try the same on your end and let us know the feedback. .
.