DM8168 network issue

Hi

Can you share the steps/application that can be used to recreate the problem.

If you could also dump the base address of the buffer(skb->data), we could then check if this happens to line up with cache line size

Also, what version of the driver/SW are you using

Regards

Sriram

Hi Sriram,

     Thanks for you reply. I am using DVRRDK_03.00.00.00.  I have uploaded my application. 

                                                                testing on our own board (4bytes of data to be sent)

According to the screenshot we could see the serial number of the adjacent two UDP packets were the same.

But when I dumped the skb data in the emac driver, I found this two packets were different.(different was correct)

modification:

static int emac_dev_xmit(struct sk_buff *skb, struct net_device *ndev)

{

    struct device *emac_dev = &ndev->dev;

    int ret_code;

    struct emac_priv *priv = netdev_priv(ndev);

    /* If no link, return */

    if (unlikely(!priv->link)) {

        if (netif_msg_tx_err(priv) && net_ratelimit())

            dev_err(emac_dev, "DaVinci EMAC: No link to transmit");

        goto fail_tx;

    }

FRAME_DUMP("TX:", skb->data, skb->len);

log：

TX:: len=46, bufp=0x8d807b02

00 0c 29 8a 91 6f 00 15 65 56 76 43 08 00 45 00 00 20 00 00 40 00 40 11 ... 00 0c 38 a1 33 38 31 37

RX:: len=60, bufp=0x8cd65064

22 b8 27 0f 00 0c 38 a1 33 38 31 37 00 00 00 00 00 00 00 00 00 00 00 00 ... 41 41 41 00 20 46 4a 45

TX:: len=46, bufp=0x8d807b02

00 0c 29 8a 91 6f 00 15 65 56 76 43 08 00 45 00 00 20 00 00 40 00 40 11 ... 00 0c 38 a0 33 38 31 38

We could see this two packets  were different.

 Any others experiencing this problem?

Hi

Your previous post seems to show different results

1.The destination MAC address field seems to be different between the frames that you are comparing

2. Only 4 bytes of data seem to be sent (as against 21 earlier)

Can you confirm if you had changed your application -and why the Destination address is different now

If you are working with ethernet ports(both being connected to the same subnet), can you re-run the test with just one port being connected to the network

Regards

Sriram

Hi Sriram,

   Sorry, I have not described clearly.  My previous post is the test results on the Ti EVM board and there are 21bytes of data to be sent. My second post is the test results on our own board. I just sent 4bytes of data. So this two boards have different Destination address. Both of this two board are all have this issue.

Hi

I was actually referring to the last posting , where you were comparing between 2 frames(both with 4bytes data, with wireshark frame numbers 13258 and 13262 respectively) but the destination MAC address between the 2 frames were not the same - have you explored why this could be?

Regards

Sriram

Hi

       The destination MAC address(80:8d:xx:xx:xx) is belong to our gateway. That packet was sent to the gateway. It looks  strange.So I have made a new test.(100bytes of data to be sent)

 Any others experiencing this problem?

 Any others experiencing this problem?

Hi!

I experienced the same issue and it's now corrected.

I had to modify linux kernel drivers : davinci_emac (emac_dev_xmit) and davinci_cpdma (all the file).

Please find attached these files corrected.

Regards6560.net.zip