How To Using FMC Write?

[Dlak]

I encountered a problem when trying to send more than 5 bytes of data using the FMC.

// impossible example --

#define SD_DATA (*(volatile unsigned char *)0x60000001)

SD_DATA = 0x01;
SD_DATA = 0x02;
SD_DATA = 0x03;

 

Using this method, data is not transmitted to the same address without a delay.

However, it is possible to write data to the FMC if I change the address each time I transmit or use a delay.

 

// possible example -- 1

#define SD_DATA (*(volatile unsigned char *)0x60000001)

SD_DATA = 0x01;
HAL_Delay(1);
SD_DATA = 0x02;
HAL_Delay(1);
SD_DATA = 0x03;

// possible example -- 2

#define SD_DATA1 (*(volatile unsigned char *)0x60000001)
#define SD_DATA2 (*(volatile unsigned char *)0x60000002)
#define SD_DATA3 (*(volatile unsigned char *)0x60000003)
SD_DATA1 = 0x01;
SD_DATA2 = 0x02;
SD_DATA3 = 0x03;

 

What could be the problem?

 

We used the STM32F756, and the following is the FMC initialization.

static void MX_FMC_Init(void)
{
 FMC_NORSRAM_TimingTypeDef Timing;


 /** Perform the NOR1 memory initialization sequence
 */
 hnor1.Instance = FMC_NORSRAM_DEVICE;
 hnor1.Extended = FMC_NORSRAM_EXTENDED_DEVICE;
 /* hnor1.Init */
 hnor1.Init.NSBank = FMC_NORSRAM_BANK1;
 hnor1.Init.DataAddressMux = FMC_DATA_ADDRESS_MUX_DISABLE;
 hnor1.Init.MemoryType = FMC_MEMORY_TYPE_NOR;
 hnor1.Init.MemoryDataWidth = FMC_NORSRAM_MEM_BUS_WIDTH_8;
 hnor1.Init.BurstAccessMode = FMC_BURST_ACCESS_MODE_DISABLE;
 hnor1.Init.WaitSignalPolarity = FMC_WAIT_SIGNAL_POLARITY_LOW;
 hnor1.Init.WaitSignalActive = FMC_WAIT_TIMING_BEFORE_WS;
 hnor1.Init.WriteOperation = FMC_WRITE_OPERATION_ENABLE;
 hnor1.Init.WaitSignal = FMC_WAIT_SIGNAL_DISABLE;
 hnor1.Init.ExtendedMode = FMC_EXTENDED_MODE_DISABLE;
 hnor1.Init.AsynchronousWait = FMC_ASYNCHRONOUS_WAIT_DISABLE;
 hnor1.Init.WriteBurst = FMC_WRITE_BURST_DISABLE;
 hnor1.Init.ContinuousClock = FMC_CONTINUOUS_CLOCK_SYNC_ONLY;
 hnor1.Init.WriteFifo = FMC_WRITE_FIFO_ENABLE;
 hnor1.Init.PageSize = FMC_PAGE_SIZE_NONE;

/* Timing */
 Timing.AddressSetupTime = 15;
 Timing.AddressHoldTime = 15;
 Timing.DataSetupTime = 31;
 Timing.BusTurnAroundDuration = 8;
 Timing.CLKDivision = 16;
 Timing.DataLatency = 17;
 Timing.AccessMode = FMC_ACCESS_MODE_A;
 /* ExtTiming */

 if (HAL_NOR_Init(&hnor1, &Timing, NULL) != HAL_OK)
 {
 _Error_Handler(__FILE__, __LINE__);
 }
}

[mƎALLEm]

Hello and welcome to the community.

Please use the button </> to insert your code.

Please read these recommendations on how to post a thread: https://community.st.com/t5/community-guidelines/how-to-write-your-question-to-maximize-your-chances-to-find-a/ta-p/575228

Thank you for your understanding

[waclawek.jan]

In MPU, set given area to be Device (it's Normal by default, which implies that the processor can merge accesses as it sees fit).

Alternatively, you can use the remap using SYSCFG_MEMRMP.SWP_FMC, the memory area starting at 0xC000'0000 is in MPU set as Device by default.

JW

[Dlak]

We used NE1 of the FMC to map the address in the hardware. Does this mean it needs to be remapped to 0xC0000000? What is the reason for mapping to 0xC0000000?

[waclawek.jan]

> What is the reason for mapping to 0xC0000000?

As I've said above, to have the SRAM in a MPU area tagged as Device. As I've said above, the alternative to remap is to set the SRAM area to Device in MPU.

> but a hard fault occurs

Where? Why? Do you still attempt to write to 0x6000'0000?

Else, debug the hardfault as usually.

JW

 

[Dlak]

I wrote the following code and changed all instances of 0x60000000 to 0xC0000000. However, a HardFault is occurring.

ex)

#define ZD_DATA (*(volatile unsigned char *)0xC0000000)

[waclawek.jan]

You have the FMC properly initialized and the swap performed?

> HardFault is occurring

Where? Single-step in disasm to find out the exact instruction.

JW

 

[Dlak]

Upon running, a HardFault occurs immediately.

If I remove the line

SYSCFG->MEMRMP |= SYSCFG_MEMRMP_SWP_FMC_0;

the HardFault disappears.

Even without this code, attempting to access the address 0xC0000000 results in a HardFault.

[waclawek.jan]

> I have noticed that using BANK1 in FMC_Init causes a hard fault at addresses starting from 0x6000.0000.

After remap, you should not use 0x6000'0000 anymore.

JW

[Dlak]

I used 0x6000.0000 before remapping, but if BANK2 is not used during the initialization process, a hard fault occurs. What could be the problem?

[waclawek.jan]

At this point, I am already lost in what exactly do you do.

Please give us a short summary, what is your actual hardware, what are the FMC settings, and upon which action does the fault occur.

JW

[Dlak]

Due to repeated swap errors, I resolved the issue by clearing the cache.