From: Michael McMaster <michael@codesrc.com>
Date: Sun, 11 Sep 2016 11:40:33 +0000 (+1000)
Subject: Fixed fsmc timing some more so it's stable
X-Git-Tag: v6.0.10~1
X-Git-Url: http://git.codesrc.com/gitweb.cgi?a=commitdiff_plain;h=fb9c26453038f15ec98eb0e151e23d2af38327b2;p=SCSI2SD-V6.git

Fixed fsmc timing some more so it's stable
 - fsmc interface is now 16bits
---

diff --git a/STM32CubeMX/SCSI2SD-V6/Src/fsmc.c b/STM32CubeMX/SCSI2SD-V6/Src/fsmc.c
index d01e9c33..8386762a 100755
--- a/STM32CubeMX/SCSI2SD-V6/Src/fsmc.c
+++ b/STM32CubeMX/SCSI2SD-V6/Src/fsmc.c
@@ -55,7 +55,7 @@ void MX_FSMC_Init(void)
   hsram1.Init.NSBank = FSMC_NORSRAM_BANK1;
   hsram1.Init.DataAddressMux = FSMC_DATA_ADDRESS_MUX_ENABLE;
   hsram1.Init.MemoryType = FSMC_MEMORY_TYPE_PSRAM;
-  hsram1.Init.MemoryDataWidth = FSMC_NORSRAM_MEM_BUS_WIDTH_8;
+  hsram1.Init.MemoryDataWidth = FSMC_NORSRAM_MEM_BUS_WIDTH_16;
   hsram1.Init.BurstAccessMode = FSMC_BURST_ACCESS_MODE_DISABLE;
   hsram1.Init.WaitSignalPolarity = FSMC_WAIT_SIGNAL_POLARITY_LOW;
   hsram1.Init.WrapMode = FSMC_WRAP_MODE_DISABLE;
@@ -67,18 +67,17 @@ void MX_FSMC_Init(void)
   hsram1.Init.WriteBurst = FSMC_WRITE_BURST_DISABLE;
   /* Timing */
 
-  // 1 clock to read the address, but since the signals aren't stable
-  // at 96MHz we wait two more clocks
-  Timing.AddressSetupTime = 3;
+  // 1 clock to read the address, + 3 for the synchroniser
+  Timing.AddressSetupTime = 4;
   Timing.AddressHoldTime = 1;
 
-  Timing.DataSetupTime = 5;
-  // +1 clock hold time, +2 clock to let the bus settle (unstable at 96MHz)
-  // +1 clock to process read, 1 clock to output
+  // 3 for synchroniser, 1 to skip hold time, 1 to process read, 1 to output
+  Timing.DataSetupTime = 6;
 
-  // Allow a clock for us to release signals, or else we'll read
-  // our own output data as an address.
-  Timing.BusTurnAroundDuration = 1;
+  // Allow a clock for us to release signals, plus 3 for the synchroniser to
+  // realise the cycle has ended. Need to avoid both devices acting as outputs
+  // on the multiplexed lines at the same time.
+  Timing.BusTurnAroundDuration = 4;
 
   Timing.CLKDivision = 16; // Ignored for async
   Timing.DataLatency = 17; // Ignored for async
@@ -108,6 +107,11 @@ static void HAL_FSMC_MspInit(void){
   PE8   ------> FSMC_DA5
   PE9   ------> FSMC_DA6
   PE10   ------> FSMC_DA7
+  PE11   ------> FSMC_DA8
+  PE12   ------> FSMC_DA9
+  PE13   ------> FSMC_DA10
+  PE14   ------> FSMC_DA11
+  PE15   ------> FSMC_DA12
   PD14   ------> FSMC_DA0
   PD15   ------> FSMC_DA1
   PD0   ------> FSMC_DA2
@@ -115,10 +119,15 @@ static void HAL_FSMC_MspInit(void){
   PD4   ------> FSMC_NOE
   PD5   ------> FSMC_NWE
   PD7   ------> FSMC_NE1
+  PD8   ------> FSMC_DA13
+  PD9   ------> FSMC_DA14
+  PD10  ------> FSMC_DA15
   PB7   ------> FSMC_NL
+  PE0   ------> FSMC_NBL0
+  PE1   ------> FSMC_NBL1
   */
   /* GPIO_InitStruct */
-  GPIO_InitStruct.Pin = GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10;
+  GPIO_InitStruct.Pin = GPIO_PIN_0|GPIO_PIN_1|GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10|GPIO_PIN_11|GPIO_PIN_12|GPIO_PIN_13|GPIO_PIN_14|GPIO_PIN_15;
   GPIO_InitStruct.Mode = GPIO_MODE_AF_PP;
   GPIO_InitStruct.Pull = GPIO_NOPULL;
   GPIO_InitStruct.Speed = GPIO_SPEED_HIGH;
@@ -128,7 +137,7 @@ static void HAL_FSMC_MspInit(void){
 
   /* GPIO_InitStruct */
   GPIO_InitStruct.Pin = GPIO_PIN_14|GPIO_PIN_15|GPIO_PIN_0|GPIO_PIN_1 
-                          |GPIO_PIN_4|GPIO_PIN_5|GPIO_PIN_7;
+                          |GPIO_PIN_4|GPIO_PIN_5|GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10;
   GPIO_InitStruct.Mode = GPIO_MODE_AF_PP;
   GPIO_InitStruct.Pull = GPIO_NOPULL;
   GPIO_InitStruct.Speed = GPIO_SPEED_HIGH;
@@ -178,6 +187,11 @@ static void HAL_FSMC_MspDeInit(void){
   PE8   ------> FSMC_DA5
   PE9   ------> FSMC_DA6
   PE10   ------> FSMC_DA7
+  PE11   ------> FSMC_DA8
+  PE12   ------> FSMC_DA9
+  PE13   ------> FSMC_DA10
+  PE14   ------> FSMC_DA11
+  PE15   ------> FSMC_DA12
   PD14   ------> FSMC_DA0
   PD15   ------> FSMC_DA1
   PD0   ------> FSMC_DA2
@@ -185,13 +199,18 @@ static void HAL_FSMC_MspDeInit(void){
   PD4   ------> FSMC_NOE
   PD5   ------> FSMC_NWE
   PD7   ------> FSMC_NE1
+  PD8   ------> FSMC_DA13
+  PD9   ------> FSMC_DA14
+  PD10  ------> FSMC_DA15
   PB7   ------> FSMC_NL
+  PE0   ------> FSMC_NBL0
+  PE1   ------> FSMC_NBL1
   */
 
-  HAL_GPIO_DeInit(GPIOE, GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10);
+  HAL_GPIO_DeInit(GPIOE, GPIO_PIN_0|GPIO_PIN_1|GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10|GPIO_PIN_11|GPIO_PIN_12|GPIO_PIN_13|GPIO_PIN_14|GPIO_PIN_15);
 
   HAL_GPIO_DeInit(GPIOD, GPIO_PIN_14|GPIO_PIN_15|GPIO_PIN_0|GPIO_PIN_1 
-                          |GPIO_PIN_4|GPIO_PIN_5|GPIO_PIN_7);
+                          |GPIO_PIN_4|GPIO_PIN_5|GPIO_PIN_7|GPIO_PIN_8|GPIO_PIN_9|GPIO_PIN_10);
 
   HAL_GPIO_DeInit(GPIOB, GPIO_PIN_7);
 
diff --git a/rtl/fpga_bitmap.o b/rtl/fpga_bitmap.o
index 5d95ad37..604983ff 100644
Binary files a/rtl/fpga_bitmap.o and b/rtl/fpga_bitmap.o differ
diff --git a/src/firmware/disk.c b/src/firmware/disk.c
index be912166..7d6d4757 100755
--- a/src/firmware/disk.c
+++ b/src/firmware/disk.c
@@ -619,13 +619,16 @@ void scsiDiskPoll()
 					dmaBytes = bytesPerSector % SD_SECTOR_SIZE;
 					if (dmaBytes == 0) dmaBytes = SD_SECTOR_SIZE;
 				}
-				for (int k = 0; k < dmaBytes; ++k)
+
+				uint16_t* scsiDmaData = (uint16_t*) &(scsiDev.data[SD_SECTOR_SIZE * (i % buffers)]);
+				for (int k = 0; k < (dmaBytes + 1) / 2; ++k)
 				{
-					scsiPhyTx(scsiDev.data[SD_SECTOR_SIZE * (i % buffers) + k]);
+					scsiPhyTx(scsiDmaData[k]);
 				}
 				i++;
 				while (!scsiPhyComplete() && !scsiDev.resetFlag) {}
 				scsiPhyFifoFlip();
+				scsiSetDataCount(dmaBytes);
 			}
 #endif
 		}
diff --git a/src/firmware/scsiPhy.c b/src/firmware/scsiPhy.c
index ba71f659..7b4e4633 100755
--- a/src/firmware/scsiPhy.c
+++ b/src/firmware/scsiPhy.c
@@ -109,8 +109,8 @@ static void assertFail()
 	}
 }
 
-static void
-startScsiRx(uint32_t count)
+void
+scsiSetDataCount(uint32_t count)
 {
 	*SCSI_DATA_CNT_HI = count >> 8;
 	*SCSI_DATA_CNT_LO = count & 0xff;
@@ -126,7 +126,7 @@ scsiReadByte(void)
 		assertFail();
 	}
 #endif
-	startScsiRx(1);
+	scsiSetDataCount(1);
 
 	trace(trace_spinPhyRxFifo);
 	while (!scsiPhyComplete() && likely(!scsiDev.resetFlag)) {}
@@ -151,9 +151,10 @@ scsiReadByte(void)
 static void
 scsiReadPIO(uint8_t* data, uint32_t count)
 {
-	for (int i = 0; i < count; ++i)
+	uint16_t* fifoData = (uint16_t*)data;
+	for (int i = 0; i < (count + 1) / 2; ++i)
 	{
-		data[i] = scsiPhyRx();
+		fifoData[i] = scsiPhyRx(); // TODO ASSUMES LITTLE ENDIAN
 	}
 	// TODO scsiDev.parityError = scsiDev.parityError || SCSI_Parity_Error_Read();
 }
@@ -168,7 +169,11 @@ scsiReadDMA(uint8_t* data, uint32_t count)
 	scsiTxDMAComplete = 1; // TODO not used much
 	scsiRxDMAComplete = 0; // TODO not used much
 
-	HAL_DMA_Start(&fsmcToMem, (uint32_t) SCSI_FIFO_DATA, (uint32_t) data, count);
+	HAL_DMA_Start(
+		&fsmcToMem,
+		(uint32_t) SCSI_FIFO_DATA,
+		(uint32_t) data,
+		(count + 1) / 2);
 }
 
 int
@@ -209,7 +214,7 @@ scsiRead(uint8_t* data, uint32_t count)
 		chunk = chunk & 0xFFFFFFF8;
 	}
 #endif
-	startScsiRx(chunk);
+	scsiSetDataCount(chunk);
 
 	while (i < count && likely(!scsiDev.resetFlag))
 	{
@@ -226,7 +231,7 @@ scsiRead(uint8_t* data, uint32_t count)
 #endif
 		if (nextChunk > 0)
 		{
-			startScsiRx(nextChunk);
+			scsiSetDataCount(nextChunk);
 		}
 
 #ifdef SCSI_FSMC_DMA
@@ -278,6 +283,8 @@ scsiWriteByte(uint8_t value)
 	scsiPhyTx(value);
 	scsiPhyFifoFlip();
 
+	scsiSetDataCount(1);
+
 	trace(trace_spinTxComplete);
 	while (!scsiPhyComplete() && likely(!scsiDev.resetFlag)) {}
 
@@ -292,9 +299,10 @@ scsiWriteByte(uint8_t value)
 static void
 scsiWritePIO(const uint8_t* data, uint32_t count)
 {
-	for (int i = 0; i < count; ++i)
+	uint16_t* fifoData = (uint16_t*)data;
+	for (int i = 0; i < (count + 1) / 2; ++i)
 	{
-		scsiPhyTx(data[i]);
+		scsiPhyTx(fifoData[i]);
 	}
 }
 
@@ -383,6 +391,7 @@ scsiWrite(const uint8_t* data, uint32_t count)
 #endif
 
 		scsiPhyFifoFlip();
+		scsiSetDataCount(chunk);
 		i += chunk;
 	}
 	while (!scsiPhyComplete() && likely(!scsiDev.resetFlag))
@@ -486,7 +495,7 @@ void scsiPhyReset()
 	*SCSI_CTRL_TIMING = SCSI_DEFAULT_TIMING;
 
 	// DMA Benchmark code
-	// Currently 6.6MB/s. Assume 13MB/s is achievable with 16 bits
+	// Currently 11MB/s.
 	#ifdef DMA_BENCHMARK
 	while(1)
 	{
@@ -551,7 +560,7 @@ void scsiPhyReset()
 			&fsmcToMem,
 			(uint32_t) SCSI_FIFO_DATA,
 			(uint32_t) &scsiDev.data[0],
-			SCSI_FIFO_DEPTH);
+			SCSI_FIFO_DEPTH / 2);
 
 		HAL_DMA_PollForTransfer(
 			&fsmcToMem,
@@ -598,11 +607,11 @@ static void scsiPhyInitDMA()
 		memToFSMC.Init.PeriphInc = DMA_PINC_ENABLE;
 		memToFSMC.Init.MemInc = DMA_MINC_DISABLE;
 		memToFSMC.Init.PeriphDataAlignment = DMA_PDATAALIGN_WORD;
-		memToFSMC.Init.MemDataAlignment = DMA_MDATAALIGN_BYTE;
+		memToFSMC.Init.MemDataAlignment = DMA_MDATAALIGN_HALFWORD;
 		memToFSMC.Init.Mode = DMA_NORMAL;
 		memToFSMC.Init.Priority = DMA_PRIORITY_LOW;
 		// FIFO mode is needed to allow conversion from 32bit words to the
-		// 8bit FSMC interface.
+		// 16bit FSMC interface.
 		memToFSMC.Init.FIFOMode = DMA_FIFOMODE_ENABLE;
 
 		// We only use 1 word (4 bytes) in the fifo at a time. Normally it's
@@ -622,7 +631,7 @@ static void scsiPhyInitDMA()
 		fsmcToMem.Init.Direction = DMA_MEMORY_TO_MEMORY;
 		fsmcToMem.Init.PeriphInc = DMA_PINC_DISABLE;
 		fsmcToMem.Init.MemInc = DMA_MINC_ENABLE;
-		fsmcToMem.Init.PeriphDataAlignment = DMA_PDATAALIGN_BYTE;
+		fsmcToMem.Init.PeriphDataAlignment = DMA_PDATAALIGN_HALFWORD;
 		fsmcToMem.Init.MemDataAlignment = DMA_MDATAALIGN_WORD;
 		fsmcToMem.Init.Mode = DMA_NORMAL;
 		fsmcToMem.Init.Priority = DMA_PRIORITY_LOW;
diff --git a/src/firmware/scsiPhy.h b/src/firmware/scsiPhy.h
index fa7e1684..718b25fb 100755
--- a/src/firmware/scsiPhy.h
+++ b/src/firmware/scsiPhy.h
@@ -18,25 +18,25 @@
 #define SCSIPHY_H
 
 #define SCSI_CTRL_IDMASK ((volatile uint8_t*)0x60000000)
-#define SCSI_CTRL_PHASE ((volatile uint8_t*)0x60000001)
-#define SCSI_CTRL_BSY ((volatile uint8_t*)0x60000002)
-#define SCSI_FIFO_SEL ((volatile uint8_t*)0x60000003)
-#define SCSI_DATA_CNT_HI ((volatile uint8_t*)0x60000004)
-#define SCSI_DATA_CNT_LO ((volatile uint8_t*)0x60000005)
-#define SCSI_DATA_CNT_SET ((volatile uint8_t*)0x60000006)
-#define SCSI_CTRL_DBX ((volatile uint8_t*)0x60000007)
-#define SCSI_CTRL_SYNC_OFFSET ((volatile uint8_t*)0x60000008)
-#define SCSI_CTRL_TIMING ((volatile uint8_t*)0x60000009)
-#define SCSI_CTRL_TIMING2 ((volatile uint8_t*)0x6000000A)
-
-#define SCSI_STS_FIFO ((volatile uint8_t*)0x60000010)
-#define SCSI_STS_ALTFIFO ((volatile uint8_t*)0x60000011)
-#define SCSI_STS_FIFO_COMPLETE ((volatile uint8_t*)0x60000012)
-#define SCSI_STS_SELECTED ((volatile uint8_t*)0x60000013)
-#define SCSI_STS_SCSI ((volatile uint8_t*)0x60000014)
-#define SCSI_STS_DBX ((volatile uint8_t*)0x60000015)
-
-#define SCSI_FIFO_DATA ((volatile uint8_t*)0x60000020)
+#define SCSI_CTRL_PHASE ((volatile uint8_t*)0x60000002)
+#define SCSI_CTRL_BSY ((volatile uint8_t*)0x60000004)
+#define SCSI_FIFO_SEL ((volatile uint8_t*)0x60000006)
+#define SCSI_DATA_CNT_HI ((volatile uint8_t*)0x60000008)
+#define SCSI_DATA_CNT_LO ((volatile uint8_t*)0x6000000A)
+#define SCSI_DATA_CNT_SET ((volatile uint8_t*)0x6000000C)
+#define SCSI_CTRL_DBX ((volatile uint8_t*)0x6000000E)
+#define SCSI_CTRL_SYNC_OFFSET ((volatile uint8_t*)0x60000010)
+#define SCSI_CTRL_TIMING ((volatile uint8_t*)0x60000012)
+#define SCSI_CTRL_TIMING2 ((volatile uint8_t*)0x60000014)
+
+#define SCSI_STS_FIFO ((volatile uint8_t*)0x60000020)
+#define SCSI_STS_ALTFIFO ((volatile uint8_t*)0x60000022)
+#define SCSI_STS_FIFO_COMPLETE ((volatile uint8_t*)0x60000024)
+#define SCSI_STS_SELECTED ((volatile uint8_t*)0x60000026)
+#define SCSI_STS_SCSI ((volatile uint8_t*)0x60000028)
+#define SCSI_STS_DBX ((volatile uint8_t*)0x6000002A)
+
+#define SCSI_FIFO_DATA ((volatile uint16_t*)0x60000040)
 #define SCSI_FIFO_DEPTH 512
 
 
@@ -74,6 +74,8 @@ void scsiPhyReset(void);
 void scsiEnterPhase(int phase);
 void scsiEnterBusFree(void);
 
+void scsiSetDataCount(uint32_t count);
+
 void scsiWrite(const uint8_t* data, uint32_t count);
 void scsiRead(uint8_t* data, uint32_t count);
 void scsiWriteByte(uint8_t value);
@@ -82,12 +84,6 @@ uint8_t scsiReadByte(void);
 
 void sdTmpRead(uint8_t* data, uint32_t lba, int sectors);
 void sdTmpWrite(uint8_t* data, uint32_t lba, int sectors);
-#if 0
-
-// Contains the odd-parity flag for a given 8-bit value.
-extern const uint8_t Lookup_OddParity[256];
-
-#endif
 
 extern volatile uint8_t scsiRxDMAComplete;
 extern volatile uint8_t scsiTxDMAComplete;