STM32F103ZE RS485 DMA communication problem

[arsi]

Posted on October 31, 2011 at 18:54

HI!

I have a problem

with the following code

..

After

a few hours to

several days

.

The maximum

was

to

14 days

Data

transmission

stops working

.

Receive and transmit

works

in

one task

. (BACNET protocol)

For

DMA

uses internal

RAM

.

Receive

a

nd

other

tasks

work

s

further

.

void USART3_SendData(unsigned char buffer[], int Len) {

    if (Len > 500) {         Len = 500;     }     memcpy(TxBuffer3Out, buffer, Len);     DMA_InitTypeDef DMA_InitStructure5;     DMA_DeInit(DMA1_Channel2);     DMA_InitStructure5.DMA_PeripheralBaseAddr = USART3_DR_Base;     DMA_InitStructure5.DMA_MemoryBaseAddr = TxBuffer3Out;     DMA_InitStructure5.DMA_DIR = DMA_DIR_PeripheralDST;     DMA_InitStructure5.DMA_BufferSize = Len;     DMA_InitStructure5.DMA_PeripheralInc = DMA_PeripheralInc_Disable;     DMA_InitStructure5.DMA_MemoryInc = DMA_MemoryInc_Enable;     DMA_InitStructure5.DMA_PeripheralDataSize = DMA_PeripheralDataSize_Byte;     DMA_InitStructure5.DMA_MemoryDataSize = DMA_MemoryDataSize_Byte;     DMA_InitStructure5.DMA_Mode = DMA_Mode_Normal;     DMA_InitStructure5.DMA_Priority = DMA_Priority_Low;     DMA_InitStructure5.DMA_M2M = DMA_M2M_Disable;     DMA_ITConfig(DMA1_Channel2, DMA_IT_TC, ENABLE);     DMA_ITConfig(DMA1_Channel2, DMA_IT_TE, ENABLE);     DMA_Init(DMA1_Channel2, &DMA_InitStructure5);     USART_DMACmd(USART3, USART_DMAReq_Tx, ENABLE);     DMA_ClearFlag(DMA1_FLAG_TC2);     NVIC_EnableIRQ(DMA1_Channel2_IRQn);     RS485_rts_enable(true);     DMA1_Channel2->CCR |= DMA_CCR2_EN;     OS_FlagID flag;     start:     flag = CoWaitForSingleFlag(USART3txFlag, 500);     if (flag != E_OK && flag != E_TIMEOUT) {          goto start;     }     CoTimeDelay(0, 0, 0, 20);     RS485_rts_enable(false); }

void DMA1_Channel2_IRQHandler(void) {//USART3 TX

    if (DMA_GetITStatus(DMA1_IT_HT2) == SET) {         DMA_ClearITPendingBit(DMA1_IT_HT2);     } else if (DMA_GetITStatus(DMA1_IT_TC2) == SET) {         DMA_ClearITPendingBit(DMA1_IT_TC2);         USART_ITConfig(USART3, USART_IT_TC, ENABLE);     } else if (DMA_GetITStatus(DMA1_IT_TE2) == SET) {         DMA_ClearITPendingBit(DMA1_IT_TE2);     }     return; }

void USART3_IRQHandler(void) {

    if (USART_GetITStatus(USART3, USART_IT_TC) == SET) {         USART_ClearITPendingBit(USART3, USART_IT_TC);         USART_ITConfig(USART3, USART_IT_TC, DISABLE);         CoEnterISR();         isr_SetFlag(USART3txFlag);         CoExitISR();         return;     }

}

Any

idea

??

Thanks!!

ArSi

[Tesla DeLorean]

Posted on November 01, 2011 at 01:37

I'm unsure of the wisdom of constantly enabling/disabling interrupts, especially the USART TC one. And the point of having DMA interrupts if you're not actually doing anything with them.

I guess I'd attack this keeping the USART TC interrupt enabled, because it will always trail the DMA completion, and use it to progress the USART DMA to the next. Consider a method of chaining the pending data so it can transition more seamlessly between blocks of DMA'd data. Is all data on the USART sent via DMA? In terms of debugging the current code, and the failures, you'll need to examine what's not getting signalled, and looking at the USART and DMA registers in the failing condition to see if they are indicative of anything. For example the USART is holding an error state precluding further rx/tx operation until cleared. DMA can be very fussy about sequencing.

[arsi]

Posted on November 01, 2011 at 10:46

The

same code

I have

used for

usart1

,

usart2

.

This

is

just

a

RS232

,

problem does not occur

.

Quantity

of data

is

transmitted

mainly through

usart1

significantly larger

.

When it comes to

DMA

error

flag

was

there

so I

included

the command

to

restart

the processor

, but

did not come

.

Data

sent

completely

and

without problems.

Debugger

I

was

hanging

there

already

3 days

,

but as

it happens

error

had not occurred

.

Most

annoying

,

there are

lots

of

complex

code.

And it

works.

And

does not

thing

which should

run without

problems

[Tesla DeLorean]

Posted on November 01, 2011 at 15:03

there are

lots

of

complex

code.

Indeed, perhaps overly so, I'd be worried about your race condition and priority inversion problems, based on the code fragment presented.