MCU-PLUS-SDK-AM243X: AM243x LaunchPad - Ethernet/IP adapter generic device mii

Hi Stefano,

I am unable to access the Wireshark log that you shared. Can you please share it again for our analysis?

Regards
Archit Dev

Hi, I have uploaded log of wireshark as answer from email, because here i can not upload, i don't know why

Hi Stefano,

If possible, can you zip your wireshark log and share the zip file here? 

Regards
Archit Dev

demo_ti.zip

Hi Archit Dev,

Here you can find zip with log of wireshark.

Hi Stefano,

the reason is, as the session was stablished in ForwardOpen request, the master has set the Connection Timeout Multiplier to 8, which means, if the device does not receive message from master after this time period, it will conclude that a timeout has happened and close the connection.

Best regards
Pourya 

Ok, thank you. 

In your opinion, this option can be modified in eds file or in other way?

In order to have a exchange between master and slave continuosly

Hi Stefano,

it depends on the tool you are using, for example if you are using a PLC it should be easily configurable.
The important question is, why does your master stops sending I/O data to keep the communication alive? May I ask which tool you are using?
And also, one more issue I see, is that the master does not use port 2222 for CIP I/O data, but it should actually according to specification.

Best regards
Pourya

Hi,

I'm using this SW: https://www.bihl-wiedemann.de/it/prodotti/software/simulatore-master-ethernetip on my PC.

But from my side, it seems that's slave use wrong port for CIP I/O. 

For example, in the first reply, src port of slave is not 2222.

Thank you

Hi Stefano,
My apologies, yes you are right in the Wireshark the device is using dynamic-range Source port ID's The device IP-Address is 192.168.1.10, correct? But this issue is resolved in SDK 09_02_00_15. Can you confirm that you are using SDK 09.02.00.15?
Regarding the tool, I haven't used that tool so I can't provide input as to why it stops the communication. You could consult the Software provider.
You could also use Hilscher Ethernet/IP tool for evaluation purpose, please note the Copyright notice.

Best regards
Pourya

Hi, 

Yes, I'm using sdk 09.02.00.15 and the ip is 192.168.1.10

As attachment you can find log of compilation and what the terminal prints during execution

8512.log.zip

Hi Stefano,

I double checked, the fix related to source port number is for next release, sorry for the confusion. But nevertheless, that shouldn't make much of trouble, unless the software you are using strictly expect CIP I/O data from port 2222.
Have you tried with other tools?

Best regards
Pourya

Hi,

I have another question.

Can you give me an example of use of callback function about change of data during cyclic exchange?

In particular, I think that when I change value of attribute of instance a callback function is triggered by stack.

Can you help me about that? Always referred to generic_device example

For example, as you can see in this screen I have created class, instance and attribute, but when I change the value of attribute callback are not triggered.

I see that the attribute changes its value both with debugger using EI_API_CIP_getAttr_dint function for example and with Hilsher Ethernet/IP tool reading attributes.

Hi Pourya,

let me say if I have to open another thread or if you can answer my last question in this thread.

Hi Stefano,

when the attribute is accessed from remote side (for example from your Ethernet/IP Tool), then the callback should get triggered.
However, when you access the same attribute using the internal functions, like EI_API_CIP_getAttr_dint, the callback is not triggered. 
In short, the callbacks are meant to be used only when triggered from Remote side.
Is that what you are experiencing?

Best regards
Pourya 

Yes, but it does not happen, also if comment EI_API_CIP_getAttr_dint function in run loop.

My expectation is that for example when I access in this way, callback is triggered.

May I do something wrong?

Hi Pourya,

In particular my question is about message that I send in cyclic exchange, my expectation is that if i send bytes during cyclic exchange one callback can be triggered.

Hi Stefano, 

in your code, one more call is still necessary to register your callback: EI_API_CIP_setInstanceAttrFunc . 
If you want to have callback in Cyclic I/O exchange, you should consider using EI_API_CIP_addAssemblyMemberCustomMapped when creating assembly associated with your attribute. 
Creating a "normal" assembly won't trigger your callback when accessing the linked attribute.

Let me know if it does answer your question.

Best regards
Pourya

Hi Pourya,

Yes, it works, thank you a lot.

One more question. In order to support 32 bit header run/idle, is there a particular setting or I have to increase size of assembly of 4 bytes?

Bye

Hi Stefano,

yes, in SDK 9.02.15 to support 32bit header run/Idle you should increase the assembly size 4 bytes.
In the next release, there will be another API for assemblies introduced, by which you could set the assembly format to either modeless (default) or 32bit Header. 

Best regards
Pourya

Hi Pourya,

Can you give an example how to use EI_API_CIP_setInstanceAttrFunc with callback for get/set attribute? Because the prototype of function is different between example and documentation?

My expectation about get_callback is the sequent:

void attr_get_callback(EI_API_CIP_NODE_T *pCipNode_p, uint16_t classId_p, uint16_t instanceId_p, uint16_t attrId_p, uint16_t *len_p, void *pvValue_p)
{
uint16_t x = 0x0100;
memcpy(pvValue_p, &x, 2);
*len_p = 2;

}

May I do something wrong? The callback is triggered but the value is not get and wireshark log attribute invalid value

Hi Stefano,

the correct function prototype is found also in documentation, for example for get callback:

The example given in the API documentation is misleading and we will correct that in future releases, the callback functions (both get and set) should return a uint32_t type value, 0 means successful. 
See the EI_API_CIP_SAttr_t struct for more details.

Other than that, your code seems fine, (for get you should update both length and the value, don't forget to return 0 on success).

Best regards
Pourya

Hi Pourya,

Can I have an example regarding use of DHCP? Because I set DHCP as configuration_method but it seems that does not obtain a valid IP

Hi Pourya,

I see that module obtains a valid IP address with DHCP and on Hilsher Tool the IP is correct, but in debug it remains the old IP address.

Why do I have this behaviour? Do I have to set parameter in some structure?

Hi Stefano,

there is a factory default setting which you can configure according to your needs, it is found in appRst.c:

If Reset request Type 1 is received the device will reset to this factory-default.

Can you elaborate, what you mean by "in debug it remains the old IP address"?

Best regards
Pourya

Hi Pourya,

Yes, but if I set DHCP as configuration method I see that:

- 

192.168.1.3 is IP Address of DHCP server, while 192.168.1.4 is IP address of Ethernet/IP module

On Hilsher Tool, I have this configuration and it's good

But in debug, for example, with this function ipAddr is 0

For using DHCP, I have modified the example in this way:

EI_APP_RST_adapterFactoryDefaultValues2 is equal to your Image

Hi Pourya,

Also on the example if I enable DHCP on launchpad, i see that IP is 192.168.1.5(different MAC address from my official board)

But this variable is always equal to 192.168.1.10(default IP address)

Hi Stefano,

when you change your IP settings and do a normal reset / Power-cycle, the last changes will be used and not the "factory default", the default setting will only be used when you do factory reset via Reset Service (type 1) or the non-volatile memory containing latest changes were corrupt, so the factory default should be applied.

Regarding EI_API_ADP_getIpAddr function not returning valid value in case of DHCP, you are correct, we will track this issue.

Best regards
Pourya

Hi Pourya,

Regarding issue of  EI_API_ADP_getIpAddr function, could you inform me when the issue will be resolved? Because I'm developing a new module and this function is very important.

Thank you

Best regards.

Stefano

Hi Stefano,

Sorry for delayed response. The issue is in our tracking list and with high probability it will be addressed by the next release.
I can't however provide a date, when the next release will be done.
 
Best regards,
Pourya

Hi Pourya,

Ok, thank you.

Just a question. What are the differences between using EI_API_CIP_setAttr_word in a for(;;) cycle or callback as explained in previous question for cycling data exchange? Except that using callback you have to build buffer in order way to have instance and attribute in correct position.

Best regards,

Stefano

Hi Stefano,

using the callback method would be more suited for when you can quickly provide the payload without blocking in the callback (fast/synchronous response). Using EI_API_CIP_setAttr_xxx APIs would be more suited for asynchronous change of your cyclic data, for example in multi-tasking scenarios.
Please note that, when using callback method, do not run heavy code in the callback, that would block the run time of stack and will cause instability.

Best regards,
Pourya

Hi Pourya,

Regarding datas that is sent during cycling exchange from master to slave, does we have to use EI_API_CIP_addAssemblyMemberCustomMapped to obtain the data or is there another method?

Best regards,

Stefano

Hi Stefano,

you should use EI_API_CIP_addAssemblyMemberCustomMapped only in callback method. If you want to use "polling" method, then you can use EI_API_CIP_setAttr_xxx and EI_API_CIP_getAttr_xxx API's. Please note that in "polling" method, you should use the normal EI_API_CIP_addAssemblyMember API to link the assembly to your attribute.
When using Polling method, the cyclic data exchange automatically reads/writes to the linked attributes, and you just need to collect/update them via EI_API_CIP_getAttr_xxx  and EI_API_CIP_setAttr_xxx API's.
When using callback method, you should provide/consume the data directly.

Best regards,
Pourya

Hi Pourya,

Regarding acyclic set/get with callback on attribute, is there a way to get/set an attribute with callback without create class and instances? because I have a large set to data and when I create class and instance I get EI_API_CIP_eERR_DICTIONARY_FULL error. 

Is there a limit regarding classes and instances that I can create?

Best regards

Stefano

Hi Stefano,

an attribute belongs either to a class or an instance of a class, it cannot exist as stand-alone entity. You can however increase the dictionary size in initialization phase, in app.c, in main function you can see this line:

Per default the dictionary size is 256, you can increase it according to your need, as long as you have enough RAM. Please note that, around 70 of this CIP objects are created and used internally by the stack. So maxInstanceNum - 70 is the actual number of objects available for you.

Best regards,
Pourya

Hi Pourya,

Regarding Redundancy (PRP/HSR), do you have some details of TI stack that you can suggest me to implement this feature?

In particular regarding DLR.

Best regards

Hi,

Thank you a lot. 

Now I have a problem to reach ethernet/ip device from another port. 

On my device I have two PHY: 

- DP83826E (RJ45)

- TJ1100 (100 Base T1)

If I try to reach my board through RJ45 port with DP83826E  it works properly, while if I try to reach module from TJ1100, module does not answer.

Both of PHY are detected from stack as log shown and all works fine, but i can not reach module with other port:

Do you have some suggestion?

Hi Stefano,

What is the test that you're trying to do? Is ping/EIP Explicit Messaging? What is the speed and duplex settings configured on the TJ1100 PHY?

The following is the link to the ICSS Debug Guide, that covers the most obvious use cases and debug scenarios encountered while using ICSS-EMAC : https://software-dl.ti.com/processor-industrial-sw/esd/ind_comms_sdk/am243x/09_02_00_15/docs/api_guide_am243x/ICSS_EMAC_DEBUG_GUIDE.html 

Can you use this guide to check if the Link Status/Receive/Transmit (in this order) are working as expected on that PHY? This can help us point out at which level the problem exists? 

Regards
Archit Dev

Hi,

It seems that no traffic is detected by module from TJ1100 PHY, like ping or EIP messaging. 

As first thing I can see that from dissasembly if I read registers of DP83826, PRU_PHY_Command function is called, while with TJ1100 it does not happen. 

Now, I will try with your suggestions.

Regards

Stefano

Hi Stefano,

Thanks for the update. 

Please let us know about your findings.

Regards
Archit Dev

Hi,

As first point of guide that you've shared I've this problem:

Hi Stefano,

Can you please try the following:

"It is possible that the link interrupt is configured incorrectly or MDIO is not triggering the interrupts. In which case one needs to put a breakpoint in the ISR ICSS_EMAC_linkISR and disconnect or connect the cable on any one port. The breakpoint should get hit, if it does not then there is some issue with the interrupt configuration."

Regards
Archit Dev

Hi,

Ok, it seems that this interrupt is not triggered putting breakpoint in disassembly.

Regards.

Stefano

Hi Stefano,

This points to potential issue in the MDIO configuration.

Are you using the MDIO Manual Mode?

Regards
Archit Dev

Hi Stefano,

As a follow-up question, can you please tell me the ICSSG instance that you're using?

Regards
Archit Dev

Hi,

I'm using ICCS0 and MDIO manual mode.