Enabling TraceX in my NetXDuo application causes `_nx_trace_event_insert' compile errors.

[MHoop.1]

I am attempting to use TraceX in my NetXDuo application. But when I enable TraceX in the Software Packs of CubeIDE (TX_ENABLE_EVENT_TRACE) I get compiler errors in all of the calls to save a tracex event. The macros that defines the calls to tracex services appear in nx_api.h, but are not defined, and hence the _nx_trace_event_ service is never defined.

 

 

I saw a discussion of this same issue over in the Renesis community where a workaound fixed the compiler issue.

https://community.renesas.com/mcu-mpu/embedded-system-platform/f/forum/10314/undefined-reference-to-nx_trace_in_line_insert-when-arp-probe

But t his workaround did not fix the STM code.

Can anyone suggest something that may help?

Thanks,

MikeH

[STM32_ZA]

Hi MikeH,

I haven't reproduced your issue can you please provide more details on:

• STM32 device
• X-CUBE-AZRTOS version

By the way can you check that the NetXDuo trace files (nx_trace_*) are added to your project when enabling TraceX support

Regards

[MHoop.1]

ZA,

Device STM32F723ZGTx

Board: NUCLEO-H723ZG

X-CUBE-AZRTOS Versions

My program is based off of the Nx_TCP_Echo_Server example. I have added additional user code which forwards commands received from a TCP client to USART2. The response received from the USART is then forwarded back to the client. This all works as expected, except that occasionally the network connection is dropped. I need to use TraceX to debug this issue.

By the way can you check that the NetXDuo trace files (nx_trace_*) are added to your project when enabling TraceX support

I don't see any nx_trace_ code inserted into main.c, app_nextduo.c, or app_azure_rtos.c. I manually inserted the following into app_threadx.c per the training video.

TX_THREAD thread_ptr;
#define TRACEX_BUFFER_SIZE 64000
uint8_t tracex_buffer[64000] __attribute__ ((section (".trace")));
/* USER CODE END PV */
 
...and
  tx_trace_enable(&tracex_buffer,TRACEX_BUFFER_SIZE,30);
  /* USER CODE END App_ThreadX_Init */

Thanks for the help.

MikeH

[STM32_ZA]

Hi,

Can you try with X-CUBE-AZRTOS-H7 V3.0.0

I attached an ioc file created with STM32CubeMX 6.7.0 and X-CUBE-AZRTOS-H7 V3.0.0 that helps you to enable the TraceX support for NetXDuo :

[MHoop.1]

Thanks Zouhair. I'll give this a try over the next few days and report back.

I did install V3.0 X-CUBE-AZRTOS-H7 and now see the TraceX support files as shown in your screenshot.

Thanks,

MikeH

[MHoop.1]

Zouhair,

I was able to get your sample code working and was able to capture TraceX data and view it in the TraceX app. Thanks for that.

However, I did run into several issue when trying to incorporate TraceX into my application (based on the NX_TCP_Echo_Server example code).

• I ran out of byte_pool memory when allocating all of the services in my NetxDuo application. I tried increasing the memory available for allocation of byte_pool by increasing the NetXDuo memory pool size. But when I attempted to allocate greater than 0x8100 bytes, I got a hard fault. Something in the allocation code (tx_byte_pool_search.c?) seems to crash beyond 0x8100 bytes.
• The code below causes the hard fault.

#define NX_APP_MEM_POOL_SIZE                     0x8200

• CubeIDE doesn't always save a manually change value for NetXDuo memory pool size when it is changed in the CubeIDE GUI , even after Code Generation. You can see the unchanged value in app_azure_rtos_config.h, but the GUI shows the new value.
• Your sample project did not include any BSP support code. I had to manually copy the code over from another project. This was manually intensive since I had to relink several paths, as well as other changes. Is there an easy way to add BSP support (i.e. NUCLEO-H723) to a generic .ioc project like the one you sent me?

Thanks again for your help.

Mikeh

[STM32_ZA]

Please find below my comments/answers:

1-I ran out of byte_pool memory when allocating all of the services in my NetxDuo application.

As mentioned in readme file some changes are required when changing the NX_PACKET pool size or address. The NX_PACKET start address (0x24048000) used in this example allows only pool size < 32KB so you you need to set/update start address as follows:

• STM32H723ZGTX_FLASH.ld.file

   .tcp_sec (NOLOAD) : {

  . = ABSOLUTE(0x24030000);

  *(.RxDecripSection)

  . = ABSOLUTE(0x24030060);

  *(.TxDecripSection)

 } >RAM_D1 AT> FLASH

.nx_data 0x24030200 (NOLOAD):

 {

  *(.NetXPoolSection)

 } >RAM_D1 AT >FLASH

• main.c file:

 MPU_InitStruct.Number = MPU_REGION_NUMBER1;

 MPU_InitStruct.BaseAddress = 0x24030000;

 MPU_InitStruct.Size = MPU_REGION_SIZE_64KB;

 MPU_InitStruct.SubRegionDisable = 0x0;

 MPU_InitStruct.AccessPermission = MPU_REGION_FULL_ACCESS;

 MPU_InitStruct.DisableExec = MPU_INSTRUCTION_ACCESS_ENABLE;

 MPU_InitStruct.IsShareable = MPU_ACCESS_NOT_SHAREABLE;

 MPU_InitStruct.IsCacheable = MPU_ACCESS_CACHEABLE;

2.CubeIDE doesn't always save a manually change value for NetXDuo memory pool size when it is changed in the CubeIDE GUI , even after Code Generation

Do you have same behavior with STM32CubeMX 6.7.0 ?

3-Your sample project did not include any BSP support code

BSP support is not managed by X-CUBE-AZRTOS and BSP files should be added manually by user or using .extSettings file which allows to add paths and files to project:

• create or copy an existing .extSettings file at example root folder: You can find example here : Projects\STM32H735G-DK\Applications\USBX\Ux_Host_MSC\.extSettings
• Update .extSettings
• Generate project

[ProjectFiles]

HeaderPath=..\..\..\..\..\..\Drivers\BSP\STM32H7xx_Nucleo

[Others]

HALModule=

[Groups]

Drivers/BSP/STM32H7xx_Nucleo=../../../../../../Drivers/BSP/STM32H7xx_Nucleo/stm32h7xx_nucleo.c;

[MHoop.1]

Zuhair,

Thanks so much for your continued assistance.

I made the memory allocation modifications that you recommended. But I still get a Hard Fault when the byte_pool allocation attempts to allocate greater that 32kbytes.

Here is what I modified.

• I modified the STM32H723ZGTX_FLASH.ld.file per your recommendations above.
• I modified the .ioc GUI as below (code gets modified during code-regeneration if modified directly in main.c)

 

• When attempting to execute tx_byte_allocate as the byte_pool exceeds 32kbyte, I get a hard fault at this statement.

/* Allocate the memory for TCP server thread   */
  if (tx_byte_allocate(byte_pool, (VOID **) &pointer,2 *  DEFAULT_MEMORY_SIZE, TX_NO_WAIT) != TX_SUCCESS)

• The byte_pool->tx_byte_pool_search value is 0x24037b88 which is just below 32kbytes. During this allocation is when the hard fault is thrown.
• And here is the build analyzer showing the allocations. Is the order correct?

Could there be some other setting that limits the allocated memory to 32kbytes?

thx

MikeH

[STM32_ZA]

can you share your example if it's possible ?

[MHoop.1]

Do you have an email address that I could send it to?