atmel_spi: work-around required for new HW bug in AT91SAM9263 Rev.B SPI controller
We're working with an AT91SAM9263 Rev B in our design and I experienced
some inconsistency in spi-based touchscreen usage between our board and
the Atmel evaluation kit we have that runs on a Rev A chip.
The data was apparently delayed by 1 byte and got ridiculous data out of
the touchscreen driver, very strange. As everything looked normal in
the spi, touchscreen and dma logs, I contacted the Atmel support and
they triggered me on a new HW bug that appeared in the Rev B SPI
controller.
The problem is that the SPI controller on the Rev B needs that the
software reset is performed two times so that it's performed correctly.
Applying the patch below solves the issue on my Rev B board. I've tested
it as well on my Rev A evaluation kit and it has apparently no unwanted
side effect, things continue to work as expected.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
diff --git a/drivers/spi/atmel_spi.c b/drivers/spi/atmel_spi.c
index 02f9320..8abae4a 100644
--- a/drivers/spi/atmel_spi.c
+++ b/drivers/spi/atmel_spi.c
@@ -766,6 +766,7 @@
/* Initialize the hardware */
clk_enable(clk);
spi_writel(as, CR, SPI_BIT(SWRST));
+ spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
spi_writel(as, MR, SPI_BIT(MSTR) | SPI_BIT(MODFDIS));
spi_writel(as, PTCR, SPI_BIT(RXTDIS) | SPI_BIT(TXTDIS));
spi_writel(as, CR, SPI_BIT(SPIEN));
@@ -782,6 +783,7 @@
out_reset_hw:
spi_writel(as, CR, SPI_BIT(SWRST));
+ spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
clk_disable(clk);
free_irq(irq, master);
out_unmap_regs:
@@ -805,6 +807,7 @@
spin_lock_irq(&as->lock);
as->stopping = 1;
spi_writel(as, CR, SPI_BIT(SWRST));
+ spi_writel(as, CR, SPI_BIT(SWRST)); /* AT91SAM9263 Rev B workaround */
spi_readl(as, SR);
spin_unlock_irq(&as->lock);