RTOS/AM5728: DSP directed message

Hi,

For 1000Mbps , you may look at pdk_am57xx_1_0_xx\packages\ti\board\diag\gmac\src\gmac_main.c, function LinkSetup(). For how to select which port to send out packet, I will check and get back to you.

Regards, Eric

lding said:

Hi,

For 1000Mbps , you may look at pdk_am57xx_1_0_xx\packages\ti\board\diag\gmac\src\gmac_main.c, function LinkSetup(). For how to select which port to send out packet, I will check and get back to you.

Regards, Eric

Thanks a lot! I'm looking forward to your answer regarding the issue regarding the directed messages.

Hi,

For selecting the port to send packets, I thought what you did in EMAC is the same way what we did in NIMU driver. It worked in NIMU and so is expected to work in EMAC as well.

Can you upload the files you changed in the EMAC loopback test example and I can try to reproduce?

Regards, Eric

lding said:

For selecting the port to send packets, I thought what you did in EMAC is the same way what we did in NIMU driver. It worked in NIMU and so is expected to work in EMAC as well.

Yes, that specific line is actually from the NIMU driver

lding said:

Can you upload the files you changed in the EMAC loopback test example and I can try to reproduce?

Of course. Here they are

EMAC_BasicExample_evmAM572x_ibox.tar.gz

Here is my example code. The changes to the basic example are:

1. For starting the link call emac_check_status() instead of emac_check_loopback_status(). That does an emac_poll for the link negotiation.

2. Disable receive task.

3. Added the "p_pkt_desc->Flags |=  1 << EMAC_PKT_FLAG_TO_PORT_SHIFT;" line in void app_test_task_send_pkts()

Regards,

Andri

Any news? Did you succeed in reproducing the issue?

Regards,
Andri

Hi,

Here is what I did: I used processor SDK RTOS 5.2 (pdk_am57xx_1_0_13), the code is somewhat different from yours, not sure what release you used. But the code to change loopback to non-loopback is isolated, I was able to merge your changes into the test case.

The AM572x GP EVM has two Ethernet ports, the top one is ETH0, the bottom one is ETH1. In your emac_check_status(), you called the emac_poll() for both ports 0 and 1, to pass that I need two separate Ethernet connection to two separate network which I don't have.

I commented out either emac_poll (0) or emac_poll(1), based on which Ethernet port is actually connected to pass this link_status check. For this test, I connected to port 1 (bottom one) and with the code p_pkt_desc->Flags |= 1 << EMAC_PKT_FLAG_TO_PORT_SHIFT; // Send on port 0. This code expects to send out packets on port 0 only, however I was able to capture packets on port 1. The packet is defined by test_pkt[] inside test_utils.c.

So I thought I was able to reproduce your issue. I opened a ticket for engineer team for investigation.

Regards, Eric

lding said:

Hi,

Here is what I did: I used processor SDK RTOS 5.2 (pdk_am57xx_1_0_13), the code is somewhat different from yours, not sure what release you used. But the code to change loopback to non-loopback is isolated, I was able to merge your changes into the test case.

The AM572x GP EVM has two Ethernet ports, the top one is ETH0, the bottom one is ETH1. In your emac_check_status(), you called the emac_poll() for both ports 0 and 1, to pass that I need two separate Ethernet connection to two separate network which I don't have.

I commented out either emac_poll (0) or emac_poll(1), based on which Ethernet port is actually connected to pass this link_status check. For this test, I connected to port 1 (bottom one) and with the code p_pkt_desc->Flags |= 1 << EMAC_PKT_FLAG_TO_PORT_SHIFT; // Send on port 0. This code expects to send out packets on port 0 only, however I was able to capture packets on port 1. The packet is defined by test_pkt[] inside test_utils.c.

So I thought I was able to reproduce your issue. I opened a ticket for engineer team for investigation.

Regards, Eric

Thanks a lot for your update!

Hi,

Is there any update on the ticket?

Regards,
Andri

Hi,

Sorry, our engineer is still looking into the issue and I will update once they have any progress.

Regards, Eric

Eric,

From customer:

I am not yet sure if it isn’t just some configuration issue.

As far as I understand the Ethernet driver has 2 different modes. Namely dual EMAC mode and switch mode. To me it feels like the driver is just in the wrong mode, but I couldn’t find any documentation how to configure in the low level driver mode.

 

From your reply it sounded like for the NDK stack everything is working fine, so I am kind of confused why it shouldn’t work for the low level driver.

 

So my question: Is it also broken for the NDK stack and the whole dual EMAC mode or is it just a configuration issue?

Rogerio,

Please let customer know that, the dual EMAC NDK example is tested in a way that only one Ethernet cable is connected to a PC host. Then the PC is re-configured to a different subnet and connected to a different EMAC port. Certainly this example worked, but it can't rule out the AM57x sends packets to both ports (we just don't know it).

The investigation for this issue is still preliminary, it is not sure if it broke in NDK or EMAC driver. The planned fix is in the coming 6.0. We will have more clarity as time goes by.

Regards, Eric