--- a/include/linux/spi/spi.h	2009-12-04 07:00:07.000000000 +0100
+++ b/include/linux/spi/spi.h	2011-01-18 14:30:48.919450001 +0100
@@ -68,6 +68,7 @@
 	struct device		dev;
 	struct spi_master	*master;
 	u32			max_speed_hz;
+	u16			delay_usecs;
 	u8			chip_select;
 	u8			mode;
 #define	SPI_CPHA	0x01			/* clock phase */
--- a/include/linux/spi/spidev.h	2009-12-04 07:00:07.000000000 +0100
+++ b/include/linux/spi/spidev.h	2011-01-18 14:09:47.719449999 +0100
@@ -124,6 +124,8 @@
 #define SPI_IOC_RD_MAX_SPEED_HZ		_IOR(SPI_IOC_MAGIC, 4, __u32)
 #define SPI_IOC_WR_MAX_SPEED_HZ		_IOW(SPI_IOC_MAGIC, 4, __u32)
 
-
+/* Read / Write SPI device default delay us */
+#define SPI_IOC_RD_DELAY_US		_IOR(SPI_IOC_MAGIC, 5, __u32)
+#define SPI_IOC_WR_DELAY_US		_IOW(SPI_IOC_MAGIC, 5, __u32)
 
 #endif /* SPIDEV_H */
--- a/include/linux/spi/spi_bitbang.h	2009-12-04 07:00:07.000000000 +0100
+++ b/include/linux/spi/spi_bitbang.h	2011-01-18 13:33:26.731450001 +0100
@@ -50,7 +50,7 @@
 
 	/* txrx_word[SPI_MODE_*]() just looks like a shift register */
 	u32	(*txrx_word[4])(struct spi_device *spi,
-			unsigned nsecs,
+			unsigned bit_delay, unsigned byte_delay,
 			u32 word, u8 bits);
 };
 
@@ -102,20 +102,22 @@
 
 static inline u32
 bitbang_txrx_be_cpha0(struct spi_device *spi,
-		unsigned nsecs, unsigned cpol,
-		u32 word, u8 bits)
+	unsigned bit_delay, unsigned byte_delay,
+	unsigned cpol, u32 word, u8 bits)
 {
 	/* if (cpol == 0) this is SPI_MODE_0; else this is SPI_MODE_2 */
 
+	spidelay(byte_delay);
+
 	/* clock starts at inactive polarity */
 	for (word <<= (32 - bits); likely(bits); bits--) {
 
 		/* setup MSB (to slave) on trailing edge */
 		setmosi(spi, word & (1 << 31));
-		spidelay(nsecs);	/* T(setup) */
+		spidelay(bit_delay);      /* T(setup) */
 
 		setsck(spi, !cpol);
-		spidelay(nsecs);
+		spidelay(bit_delay);
 
 		/* sample MSB (from slave) on leading edge */
 		word <<= 1;
@@ -127,21 +129,23 @@
 
 static inline u32
 bitbang_txrx_be_cpha1(struct spi_device *spi,
-		unsigned nsecs, unsigned cpol,
-		u32 word, u8 bits)
+		unsigned bit_delay, unsigned byte_delay,
+		unsigned cpol, u32 word, u8 bits)
 {
 	/* if (cpol == 0) this is SPI_MODE_1; else this is SPI_MODE_3 */
 
+	spidelay(byte_delay);
+
 	/* clock starts at inactive polarity */
 	for (word <<= (32 - bits); likely(bits); bits--) {
 
 		/* setup MSB (to slave) on leading edge */
 		setsck(spi, !cpol);
 		setmosi(spi, word & (1 << 31));
-		spidelay(nsecs); /* T(setup) */
+		spidelay(bit_delay); /* T(setup) */
 
 		setsck(spi, cpol);
-		spidelay(nsecs);
+		spidelay(bit_delay);
 
 		/* sample MSB (from slave) on trailing edge */
 		word <<= 1;
--- a/drivers/spi/spidev.c	2009-12-04 07:00:07.000000000 +0100
+++ b/drivers/spi/spidev.c	2011-01-18 14:15:00.971449999 +0100
@@ -362,6 +362,9 @@
 	case SPI_IOC_RD_MAX_SPEED_HZ:
 		retval = __put_user(spi->max_speed_hz, (__u32 __user *)arg);
 		break;
+	case SPI_IOC_RD_DELAY_US:
+		retval = __put_user(spi->delay_usecs, (__u32 __user *)arg);
+		break;
 
 	/* write requests */
 	case SPI_IOC_WR_MODE:
@@ -426,7 +429,19 @@
 				dev_dbg(&spi->dev, "%d Hz (max)\n", tmp);
 		}
 		break;
+	case SPI_IOC_WR_DELAY_US:
+		retval = __get_user(tmp, (__u32 __user *)arg);
+		if (retval == 0) {
+			u32	save = spi->delay_usecs;
 
+			spi->delay_usecs = tmp;
+			retval = spi_setup(spi);
+			if (retval < 0)
+				spi->delay_usecs = save;
+			else
+				dev_dbg(&spi->dev, "%d us delay\n", tmp);
+		}
+		break;
 	default:
 		/* segmented and/or full-duplex I/O request */
 		if (_IOC_NR(cmd) != _IOC_NR(SPI_IOC_MESSAGE(0))
--- a/drivers/spi/spi_bitbang.c	2009-12-04 07:00:07.000000000 +0100
+++ b/drivers/spi/spi_bitbang.c	2011-01-18 20:53:42.079449999 +0100
@@ -49,12 +49,14 @@
 
 struct spi_bitbang_cs {
 	unsigned	nsecs;	/* (clock cycle time)/2 */
-	u32		(*txrx_word)(struct spi_device *spi, unsigned nsecs,
+	u32		(*txrx_word)(struct spi_device *spi,
+					unsigned bit_delay, unsigned byte_delay,
 					u32 word, u8 bits);
 	unsigned	(*txrx_bufs)(struct spi_device *,
 					u32 (*txrx_word)(
 						struct spi_device *spi,
-						unsigned nsecs,
+						unsigned bit_delay,
+						unsigned byte_delay,
 						u32 word, u8 bits),
 					unsigned, struct spi_transfer *);
 };
@@ -62,26 +64,42 @@
 static unsigned bitbang_txrx_8(
 	struct spi_device	*spi,
 	u32			(*txrx_word)(struct spi_device *spi,
-					unsigned nsecs,
+					unsigned bit_delay, unsigned byte_delay,
 					u32 word, u8 bits),
-	unsigned		ns,
+	unsigned		bit_delay,
 	struct spi_transfer	*t
 ) {
 	unsigned		bits = spi->bits_per_word;
 	unsigned		count = t->len;
 	const u8		*tx = t->tx_buf;
 	u8			*rx = t->rx_buf;
+	unsigned		byte_delay = spi->delay_usecs;
 
 	while (likely(count > 0)) {
 		u8		word = 0;
 
-		if (tx)
+		if (unlikely(tx))
 			word = *tx++;
-		word = txrx_word(spi, ns, word, bits);
-		if (rx)
+		word = txrx_word(spi, bit_delay, byte_delay, word, bits);
+		if (likely(rx)) {
+			/* If we receive a 0x00, fetch one extra byte to sync
+                           the state machine, then break out of the while loop. */
+			if (unlikely(!word)) {
+				txrx_word(spi, bit_delay, byte_delay, 0x00, bits); /* discard */
+				break;
+			}
+
 			*rx++ = word;
+		}
 		count -= 1;
 	}
+
+	if (unlikely(tx)) {
+		/* Signal the end of tx by sending two 0x00's. */
+		txrx_word(spi, bit_delay, byte_delay, 0x00, bits);
+		txrx_word(spi, bit_delay, byte_delay, 0x00, bits);
+	}
+
 	return t->len - count;
 }
 
@@ -156,10 +174,10 @@
 		bits_per_word = spi->bits_per_word;
 	if (bits_per_word <= 8)
 		cs->txrx_bufs = bitbang_txrx_8;
-	else if (bits_per_word <= 16)
+/*	else if (bits_per_word <= 16)
 		cs->txrx_bufs = bitbang_txrx_16;
 	else if (bits_per_word <= 32)
-		cs->txrx_bufs = bitbang_txrx_32;
+		cs->txrx_bufs = bitbang_txrx_32; */
 	else
 		return -EINVAL;
 
@@ -167,8 +185,8 @@
 	if (!hz)
 		hz = spi->max_speed_hz;
 	if (hz) {
-		cs->nsecs = (1000000000/2) / hz;
-		if (cs->nsecs > (MAX_UDELAY_MS * 1000 * 1000))
+		cs->nsecs = (1000000/2) / hz;
+		if (cs->nsecs > (MAX_UDELAY_MS * 1000))
 			return -EINVAL;
 	}
 
@@ -346,12 +364,6 @@
 			}
 			if (status > 0)
 				m->actual_length += status;
-			if (status != t->len) {
-				/* always report some kind of error */
-				if (status >= 0)
-					status = -EREMOTEIO;
-				break;
-			}
 			status = 0;
 
 			/* protocol tweaks before next transfer */
--- a/drivers/spi/spi_gpio.c	2010-12-14 01:02:26.673204002 +0100
+++ b/drivers/spi/spi_gpio.c	2011-01-18 13:29:17.915450000 +0100
@@ -156,27 +156,27 @@
  */
 
 static u32 spi_gpio_txrx_word_mode0(struct spi_device *spi,
-		unsigned nsecs, u32 word, u8 bits)
+		unsigned bit_delay, unsigned byte_delay, u32 word, u8 bits)
 {
-	return bitbang_txrx_be_cpha0(spi, nsecs, 0, word, bits);
+	return bitbang_txrx_be_cpha0(spi, bit_delay, byte_delay, 0, word, bits);
 }
 
 static u32 spi_gpio_txrx_word_mode1(struct spi_device *spi,
-		unsigned nsecs, u32 word, u8 bits)
+		unsigned bit_delay, unsigned byte_delay, u32 word, u8 bits)
 {
-	return bitbang_txrx_be_cpha1(spi, nsecs, 0, word, bits);
+	return bitbang_txrx_be_cpha1(spi, bit_delay, byte_delay, 0, word, bits);
 }
 
 static u32 spi_gpio_txrx_word_mode2(struct spi_device *spi,
-		unsigned nsecs, u32 word, u8 bits)
+		unsigned bit_delay, unsigned byte_delay, u32 word, u8 bits)
 {
-	return bitbang_txrx_be_cpha0(spi, nsecs, 1, word, bits);
+	return bitbang_txrx_be_cpha0(spi, bit_delay, byte_delay, 1, word, bits);
 }
 
 static u32 spi_gpio_txrx_word_mode3(struct spi_device *spi,
-		unsigned nsecs, u32 word, u8 bits)
+		unsigned bit_delay, unsigned byte_delay, u32 word, u8 bits)
 {
-	return bitbang_txrx_be_cpha1(spi, nsecs, 1, word, bits);
+	return bitbang_txrx_be_cpha1(spi, bit_delay, byte_delay, 1, word, bits);
 }
 
 /*----------------------------------------------------------------------*/