McBSP SPI on a C5409 does only 7 bits per word?

I'm trying to have the McBSP2 on our C5409 do SPI as a master to a serial eeprom.

I'm getting pretty close, but I doesn't quite work. If I send only one byte it works, but when I try to send more than one bytes, all the bytes but the first one have only 7 bit in them!??! (Even though I wait for TEMPTY. I tried TRDY too, with no luck).

I do not use DMA or IRQ, and I have the same behavior from 400kbps all the way to 2Mbps. I haven`t tried any faster.

Anybody has seem something like that before? Any leads on what I might try or check?

void spiInit(void){
	SPSA2 = SPCR2_SUB;
	SPCR22->raw = 0x0000;		/* Erase SPCR2 */
	SPCR22->bits.grst = 0;		/* go into global reset  */
	SPCR22->bits.xrst = 0;		/* put the tx into reset */

	SPSA2 = SPCR1_SUB;
	SPCR12->raw = 0x000;
	SPCR12->bits.rrst = 0;		/* Receiver in reset state */
	SPCR12->bits.clkstp = 0x2;	/* Set clock to be running only when we have transaction  */

	SPSA2 = PCR_SUB;
	PCR2->raw = 0x0000;
	PCR2->bits.fsxm = 1;		/* frame synchro from the sample rate generator */
	PCR2->bits.clkxm = 1;		/* we are master so we transmit the tx clock */
	PCR2->bits.clkrm = 1;		/* we are master so we transmit the rx clock */
	PCR2->bits.fsxp = 1;		/* frame synchro is active low  */
	PCR2->bits.clkxp = 1;		/* clock polarity is high*/

	SPSA2 = SRGR1_SUB;
	SRGR12->raw = 0x0000;
	SRGR12->bits.clkgdv = 100;	/* CPU internal clk is 100MHz /100 -> spi runs at 1MHz*/

	SPSA2 = SRGR2_SUB;
	SRGR22->raw = 0x0000;
	SRGR22->bits.clksp = 0;		/* clock polarity on raising edges edges */
	SRGR22->bits.clksm = 1;		/* sample rate generator derived from CPU clock */
	SRGR22->bits.fsgm = 0;



	SPSA2 = XCR1_SUB;
	XCR12->raw 		= 0x0000;
	XCR12->bits.xfrlen1 = 0;	/* 1 word per frame */
	XCR12->bits.xwdlen1 = 0;	/* 8 bits per words */

	SPSA2 = XCR2_SUB;
	XCR22->raw 		= 0x0000;
	XCR22->bits.xdatdly = 1;	/* one bit data delay on tx */

	SPSA2 = RCR1_SUB;
	RCR12->raw 		= 0x0000;
	RCR12->bits.rfrlen1 = 0;	/* 1 word per frame */
	RCR12->bits.rwdlen1 = 0;	/* 8 bits per words */

	SPSA2 = RCR2_SUB;
	RCR22->raw 		= 0x0000;
	RCR22->bits.rdatdly = 1;	/* one bit data delay on tx */

	asm("        NOP");

	SPSA2 = SPCR2_SUB;
	SPCR22->raw 	= 0x0000;
	SPCR22->bits.grst = 1; 		/* leaving global reset */
	SPCR22->bits.xrst = 1; 		/* leaving TX reset */

	asm("        NOP");

}

void spiTxByte(unsigned char byte){
	int empty = 1;

	/* wait for TX to be completed */
	while(empty == 1){
		SPSA2 = SPCR2_SUB;
		empty = SPCR22->bits.xempty;
	}

	/* transmit the actual data */
	DXR12 = (unsigned int)byte;


}

void spiTxBytes(unsigned char* byte, int len)
{
	int i;
	int ready = 0;
	/* wait for TX to be ready */
	while(ready == 0){
		SPSA2 = SPCR2_SUB;
		ready = SPCR22->bits.xrdy;
	}

	SPSA2 = XCR1_SUB;
	XCR12->bits.xfrlen1 = len-1;	/*  word per frame */

	for (i=0; i<len;i++){
		spiTxByte((unsigned int)(*byte));
		byte++;
	}

}