From: Mike Frysinger Date: Tue, 23 Mar 2010 17:29:06 -0400 Subject: [U-Boot] [PATCH v2] spi: add altera spi controller support In-Reply-To: <1269315379-6378-1-git-send-email-thomas@wytron.com.tw> References: <1269315379-6378-1-git-send-email-thomas@wytron.com.tw> Message-ID: <201003231729.07252.vapier@gentoo.org> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de On Monday 22 March 2010 23:36:19 Thomas Chou wrote: > +#include > +#include > +#include > +#include > +#include side note, but am i the only one who thinks nios headers in include/ is bad mojo ? > +static nios_spi_t *nios_spi = (nios_spi_t *)CONFIG_SYS_SPI_BASE; shouldnt this be based on the bus # so that you can support multiple ones at runtime ? i.e. you add a regs member to the spi slave struct and operate off that at runtime. > +void spi_init (void) i think the spaces should be consumed before the paren per LKML style > +{ > + /* empty read buffer */ > + if (readl (&nios_spi->status) & NIOS_SPI_RRDY) > + readl (&nios_spi->rxdata); > + return; > +} useless return ... > +struct spi_slave *spi_setup_slave (unsigned int bus, unsigned int cs, > + unsigned int max_hz, unsigned int mode) > +{ > + struct spi_slave *slave; > + > + slave = malloc (sizeof(struct spi_slave)); sizeof(*slave) > + if (!slave) > + return NULL; > + > + slave->bus = bus; > + slave->cs = cs; > + > + return slave; > +} shouldnt you be validating the bus/cs values here ? presumably you only have 1 bus atm, so bus != 0 is an error. and the cs is being used to write to an "unsigned slaveselect", so the range of values is at most 0...31 inclusive (assuming your hardware can actually support 32 CS's). > +int spi_claim_bus (struct spi_slave *slave) > +{ > + return 0; > +} > + > +void spi_release_bus (struct spi_slave *slave) > +{ > + return; > +} i'm pretty sure you should be programming either nios_spi->slaveselect or nios_spi->control here rather than in the spi_xfer func ... > +int spi_xfer (struct spi_slave *slave, unsigned int bitlen, const void > *dout, + void *din, unsigned long flags) > +{ > + int i, iter = bitlen >> 3; > + const uchar *txp = dout; > + uchar *rxp = din; > + uchar d; > + > + if (flags & SPI_XFER_BEGIN) { > + writel (1 << slave->cs, &nios_spi->slaveselect); > + writel (NIOS_SPI_SSO, &nios_spi->control); > + } > + > + for (i = 0; i < iter; i++) { > + writel (txp ? txp[i] : 0, &nios_spi->txdata); > + while (!(readl (&nios_spi->status) & NIOS_SPI_RRDY)) > + ; > + d = readl (&nios_spi->rxdata); > + if (rxp) > + rxp[i] = d; > + } > + if (flags & SPI_XFER_END) > + writel (0, &nios_spi->control); > + > + return 0; > +} since you only support multiples of 8 bytes here, your code should return an error when (bitlen & 0x7) your XFER_END should force the CS to go low ... is that what the control does, or is it the slaveselect ? -mike -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 836 bytes Desc: This is a digitally signed message part. Url : http://lists.denx.de/pipermail/u-boot/attachments/20100323/c7acb0fe/attachment.pgp