Ethernet protocol converter using TM4C1294

Hello Utpal

I know that there is a sample code being developed for basic Ethernet (just to work as communication interface between two devices). I am not sure on the release date for it.

I will ask the team member to post an update.

Regards

Amit

Hi Amit,

Is there an update to this?

Thanks,

Stephen

Hi Veikko,

Simple send/receive will be good enough for me. I'm only validating the hardware on my pcb.

I'll check out the Tivaware manual first.

Thanks,

Stephen

OK, you should be able to do that with raw ethernet quite nicely. You don't even need two boards for testing RX - just connect the ethernet cable to any Windows PC and you'll get *plenty* of packets (among which ARP and DHCP discovery, for example).

There is one gotcha in the example code. In the InitDescriptors function, you'll see:

g_psTxDescriptor[ui32Loop].ui32Count = DES1_TX_CTRL_SADDR_INSERT;

This commands the EMAC, when it reads the descriptor, to insert the ethernet address of the board to the outgoing frame in question. The setting is stored in the high bits of the .ui32Count field. However, in the PacketTransmit function, you have:

g_psTxDescriptor[g_ui32TxDescIndex].ui32Count = (uint32_t)i32BufLen;

Which effectively overwrites the previously applied setting. Now, you could do a |= assignment, but then you'd need to clear the previous buflen bits first. A better solution, IMHO, is to configure the EMAC to do the address insertion globally, ignoring the bits in the descriptor. In the example, in main() you have:

//
// Set MAC configuration options.
//
EMACConfigSet(EMAC0_BASE,
(EMAC_CONFIG_FULL_DUPLEX |
EMAC_CONFIG_CHECKSUM_OFFLOAD |
EMAC_CONFIG_7BYTE_PREAMBLE |
EMAC_CONFIG_IF_GAP_96BITS |
EMAC_CONFIG_USE_MACADDR0 |
EMAC_CONFIG_SA_FROM_DESCRIPTOR |
EMAC_CONFIG_BO_LIMIT_1024),
(EMAC_MODE_RX_STORE_FORWARD |
EMAC_MODE_TX_STORE_FORWARD |
EMAC_MODE_TX_THRESHOLD_64_BYTES |
EMAC_MODE_RX_THRESHOLD_64_BYTES), 0);

Change the highlighted setting to either EMAC_CONFIG_SA_INSERT or EMAC_CONFIG_SA_REPLACE. The difference between these is that with the former, the EMAC actually adds the source address bytes to the outgoing frame while the latter replaces the bytes in situ. I find the latter more intuitive, as then you will be sending and receiving frames with the actual structure.

Also another thing to note - if the automatic IP+ICMP/TCP/UDP checksum calculation-by-hardware feature is to be used, you need to ensure that the checksum fields are all zero prior to handing over the frame to the DMA. But for raw ethernet you needn't worry about that.

Hi Viekko, I need to communicate between two TM4C1294 boards using ethernet specifically with the PTP 1588 protocol, and not one board and PC communication like you have suggested above. If you could provide a sample code to send and receive messages between two boards it would be very helpful.

Hi Vandan,

Unfortunately I don't have sample code to do that. I have read about IEEE 1588 PTP, but haven't implemented it anywhere. But I can assure you it's not as simple as getting a working example code from someone...

But before you even dream of attempting to make PTP work, you should have rock-solid understanding (and working code) of the communication between one TM4C board and a PC (a "known good" communication partner). I believe I'm not alone when I suggest to take very small, very well controlled steps towards your final target.

Best regards,

Veikko

Veikko Immonen said:

I believe I'm not alone when I suggest to take very small, very well controlled steps towards your final target.

Not alone - really?     Let's see a, "show of hands?"    (2 oars fall overboard (from this KISS canoe) when hands raise)

Immediate gratification rules - small, controlled/methodical/thoughtful steps may have (passed) their (brief) time in the sun...   (although a "tuned" piano is always melodic...especially when "island-bound"...[this way too "inside" for all but 2])

We of course applaud your "KISS" direction...

Vandan,

A little more reading comprehension would serve you well.

First, my name is Veikko, not Viekko - I wouldn't blame you if you couldn't pronounce it but I do expect everyone to be able to ctrl-c, ctrl-v it correctly.

Second, the answer to your question can already be found in the thread. See my message on Apr 7th.

Best regards,

Veikko

Hi Veikko, my apologies for getting your name wrong.  Could you please elaborate on making a DIY stack.  Any links or reference document will be really helpful. Thanks

Regards, Vandan

A couple more tips that I remembered overnight; to simplify things, I didn't implement support for TCP nor IP fragmenting. This limits the use cases of the stack somewhat (quite severely in fact), but I considered the tradeoffs OK for my projected uses of the stack. Consider this if you wish to proceed creating your own stack.

Best regards,

Veikko