Dear all,
I scratch my head about this for too long with no progress, so...
I work with a LPC4337 with external SDRAM (256Mb / 32 MB) at the EMC. The external SDRAM is located at DYCS0:
/* SDRAM Address Base for DYCS0*/ #define BUFFER_BASE 0x28000000 // Start Memory address for data buffer #define BUFFER_SIZE 0x02000000 // 32 Mb = 33,554,432 byte (0x2800 0000 --> 0x2A00 0000)
I can read- and write- access the buffer with no problems.
I also successfully send the buffer data out through USB (USBD_ROM_LIBUSB):
void TxBuffer(const uint32_t* buf, uint32_t len) { //... libusbdev_QueueSendReq(buf, len); //... }
/* Queue the given buffer for transmision to USB host application. */ ErrorCode_t libusbdev_QueueSendReq(uint8_t *pBuf, uint32_t buf_len) { LUSB_CTRL_T *pUSB = (LUSB_CTRL_T *) &g_lusb; ErrorCode_t ret = ERR_FAILED; /* Check if a read request is pending */ if (pUSB->pTxBuf == 0) { /* Queue the read request */ pUSB->pTxBuf = pBuf; pUSB->txBuffLen = buf_len; USBD_API->hw->WriteEP(pUSB->hUsb, LUSB_IN_EP, pBuf, buf_len); ret = LPC_OK; } return ret; }
Now I added a SD card to store the same data in a file. Based on the SDMMC example (including FatFS):
void TxBuffer(const uint32_t* buf, uint32_t len) { //... SDMMC_AddData(buf, len); //... }
void SDMMC_AddData(uint8_t *pBuf, uint32_t buf_len) { FRESULT rc; /* Result code */ UINT bw; rc = f_write(&Fil, pBuf, buf_len, &bw); //... }
This f_write() call works just fine when the buffer is a byte array (for example writing a text string).
However, when i pass on my external memory location pointer as pBuf (just like with the USB variant), f_write() fails. I tracked it down to the "mem_cpy()" lines within f_write(). After a few bytes of copying from 0x28000000+, it just 'stops'. No hard fault - the debugger is just quiet for a few seconds, then disconnects from the target.
The FatFS mem_cpy() is a simple byte by byte copy:
/* Copy memory to memory */ static void mem_cpy (void* dst, const void* src, UINT cnt) { BYTE *d = (BYTE*)dst; const BYTE *s = (const BYTE*)src; if (cnt) { do *d++ = *s++; while (--cnt); } }
So I tried the native memcpy() with the same result (it never returns).
My next try was a DMA transfer, which also never completes (stays enabled with transfer size >0):
/* Copy memory to memory */ static void mem_cpy_DMA(uint32_t dstAddr, uint32_t srcAddr, uint32_t bytes) { while (LPC_GPDMA->CH[1].CONFIG & 0x1); // Wait for previous transfer to complete /* BUFFER - DMA CONFIGURATION */ LPC_GPDMA->CH[1].SRCADDR = srcAddr; LPC_GPDMA->CH[1].DESTADDR = dstAddr; LPC_GPDMA->CH[1].LLI = 0; // Linked List Items not used LPC_GPDMA->CH[1].CONTROL = (bytes << 0)| // Transfer size (0x0 << 12) | // Source Burst Size = 1 transfer per request (0x0 << 15) | // Destination Burst Size = 1 transfer per request (0x0 << 18) | // Source width = byte (0x0 << 21) | // Destination width = byte (0x1 << 24) | // Source AHB master = AHB master 1 (0x1 << 25) | // Destination AHB master = AHB master 1 (0x1 << 26) | // Source incremented after each transfer (0x1 << 27) | // Destination incremented after each transfer (0x0 << 28) | // PROT1: privileged access disabled (0x1 << 29) | // PROT2: bufferable access enabled (0x0 << 30) | // PROT3: cacheable access disabled (0x0L << 31); // Terminal count interrupt disabled LPC_GPDMA->CH[1].CONFIG = (0x1 << 0) | // Channel immediately enabled (0x0 << 1) | // Source peripheral = Don't care (transfer from memory) (0x0 << 6) | // Destination peripheral = Don't care (transfer to memory) (0x0 << 11) | // Flow control = Memory to memory (DMA control) (0x0 << 14) | // Interrupt error mask (0x0L << 15); // Terminal count interrupt mask }
Then I tried out of curiosity to change the external SDRAM back to a (smaller) local RAM, and all three memory copy variants work fine:
#define BUFFER_BASE 0x20000000 // Start Memory address buffer #define BUFFER_SIZE 0x00010000 // 64 kB AHB SRAM
So my big question is: Why does the memory copy fail?
(...Although the external SDRAM buffer obviously works fine with the USB code?)
(...Why does it work with a local SRAM address, but not with the external SDRAM address?)
All experiments are with single packets of a like 70-90 bytes, so memory size and speed etc.are not a factor at this point.
Any thoughts or ideas are much appreciated!
已解决! 转到解答。
OK, update... all the above options are fine.
The problems was something else: All 4 CLK's have to be used for the EMC interface (...). The problem was that the SDMMC init code (taken from the LPC SD/MMC example) configured CLK2 as the SD_CLK, which messed up the EMC. Configuring pin PC_0 as the SD_CLK solves the issue.
See also here: Re: SD\MMC issue
OK, update... all the above options are fine.
The problems was something else: All 4 CLK's have to be used for the EMC interface (...). The problem was that the SDMMC init code (taken from the LPC SD/MMC example) configured CLK2 as the SD_CLK, which messed up the EMC. Configuring pin PC_0 as the SD_CLK solves the issue.
See also here: Re: SD\MMC issue