Re: [Gumstix-users] Spidev lockup with omap-3.0-r103

[James C. <pea...@gm...>]

So more follow up on this:

If the transfers are done in PIO it gives a different error:
"RXS timed out" which is like 494 in drivers/spi/omap2_mcspi.c
this occurs when the received word flag doesn't flip in MCSPI_RXx

HOWEVER the interesting thing is when I switch to a duplex message (i.e. 
        xfer[0].rx_buf = (unsigned long) buf;
        xfer[0].len = len;
        xfer[0].tx_buf = (unsigned long) tx_buf;
        xfer[0].len = len;
        status = ioctl(dev_fd, SPI_IOC_MESSAGE(1), xfer);
)
the problem went away.

This section of the omap2-mcspi.c:
               /* RX_ONLY mode needs dummy data in TX reg */
                if (t->tx_buf == NULL)
                    __raw_writel(0, cs->base
                            + OMAP2_MCSPI_TX0);

is probably not working and might need  __raw_writel(*tx++, tx_reg); added
after it.


--
View this message in context: http://gumstix.8.n6.nabble.com/Spidev-lockup-with-omap-3-0-r103-tp4359993p4360727.html
Sent from the Gumstix mailing list archive at Nabble.com.