diff mbox

[v5,2/2] spi: sun4i: Add support for wait time between word transmissions

Message ID 1450352427-25350-3-git-send-email-mweseloh42@gmail.com (mailing list archive)
State New, archived
Headers show

Commit Message

Marcus Weseloh Dec. 17, 2015, 11:40 a.m. UTC
Modifies the sun4i SPI master driver to make use of the
"spi-word-wait-ns" property. This specific SPI controller needs 3 clock
cycles to set up the delay, which makes the minimum non-zero wait time
on this hardware 4 clock cycles.

Signed-off-by: Marcus Weseloh <mweseloh42@gmail.com>
---
 drivers/spi/spi-sun4i.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

Comments

Maxime Ripard Dec. 18, 2015, 11:16 a.m. UTC | #1
Hi,

On Thu, Dec 17, 2015 at 12:40:27PM +0100, Marcus Weseloh wrote:
> Modifies the sun4i SPI master driver to make use of the
> "spi-word-wait-ns" property. This specific SPI controller needs 3 clock
> cycles to set up the delay, which makes the minimum non-zero wait time
> on this hardware 4 clock cycles.
> 
> Signed-off-by: Marcus Weseloh <mweseloh42@gmail.com>
> ---
>  drivers/spi/spi-sun4i.c | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
> 
> diff --git a/drivers/spi/spi-sun4i.c b/drivers/spi/spi-sun4i.c
> index f60a6d6..3b4f5f4 100644
> --- a/drivers/spi/spi-sun4i.c
> +++ b/drivers/spi/spi-sun4i.c
> @@ -19,6 +19,7 @@
>  #include <linux/module.h>
>  #include <linux/platform_device.h>
>  #include <linux/pm_runtime.h>
> +#include <linux/of.h>
>  
>  #include <linux/spi/spi.h>
>  
> @@ -173,6 +174,8 @@ static int sun4i_spi_transfer_one(struct spi_master *master,
>  	unsigned int tx_len = 0;
>  	int ret = 0;
>  	u32 reg;
> +	int wait_clk = 0;
> +	int clk_ns = 0;
>  
>  	/* We don't support transfer larger than the FIFO */
>  	if (tfr->len > SUN4I_FIFO_DEPTH)
> @@ -261,6 +264,25 @@ static int sun4i_spi_transfer_one(struct spi_master *master,
>  
>  	sun4i_spi_write(sspi, SUN4I_CLK_CTL_REG, reg);
>  
> +	/*
> +	 * Setup wait time between words.
> +	 *
> +	 * Wait time is set in SPI_CLK cycles. The SPI hardware needs 3
> +	 * additional cycles to setup the wait counter, so the minimum delay
> +	 * time is 4 cycles.
> +	 */
> +	if (spi->word_wait_ns) {
> +		clk_ns = DIV_ROUND_UP(1000000000, tfr->speed_hz);

You should use the actual rate of the clock returned by clk_get_rate
(or probably just use mclk_rate).

The clock driver might round the frequency to something else than what
was set in clk_set_rate, which would make your calculation here a bit
off.

Maxime
Marcus Weseloh Dec. 20, 2015, 12:47 p.m. UTC | #2
Hi,

2015-12-18 12:16 GMT+01:00 Maxime Ripard <maxime.ripard@free-electrons.com>:

>>       sun4i_spi_write(sspi, SUN4I_CLK_CTL_REG, reg);
>>
>> +     /*
>> +      * Setup wait time between words.
>> +      *
>> +      * Wait time is set in SPI_CLK cycles. The SPI hardware needs 3
>> +      * additional cycles to setup the wait counter, so the minimum delay
>> +      * time is 4 cycles.
>> +      */
>> +     if (spi->word_wait_ns) {
>> +             clk_ns = DIV_ROUND_UP(1000000000, tfr->speed_hz);
>
> You should use the actual rate of the clock returned by clk_get_rate
> (or probably just use mclk_rate).
>
> The clock driver might round the frequency to something else than what
> was set in clk_set_rate, which would make your calculation here a bit
> off.

Yes, good point! And as the wait clock counter is based on the actual
SPI_CLK and not the mod clock, I need to calculate the exact clock
myself before handling the wait clock setting. Will amend the patch
and send a new version.

While looking into this, I also noticed a problem with a previous
patch of mine, which changed the spi-sun[46]i to use
transfer->speed_hz instead of the spi->max_speed_hz: I also changed
the mclk_rate calculation to be based on tfr->speed_hz, which should
have stayed with spi->max_speed_hz. In the current state, the clock
calculations only ever increase mclk_rate, wich leads to very
different clocks being set depending on which clock was used on the
previous transfer. Will send a fix for that as well in a separate
patch.

Cheers,

   Marcus
diff mbox

Patch

diff --git a/drivers/spi/spi-sun4i.c b/drivers/spi/spi-sun4i.c
index f60a6d6..3b4f5f4 100644
--- a/drivers/spi/spi-sun4i.c
+++ b/drivers/spi/spi-sun4i.c
@@ -19,6 +19,7 @@ 
 #include <linux/module.h>
 #include <linux/platform_device.h>
 #include <linux/pm_runtime.h>
+#include <linux/of.h>
 
 #include <linux/spi/spi.h>
 
@@ -173,6 +174,8 @@  static int sun4i_spi_transfer_one(struct spi_master *master,
 	unsigned int tx_len = 0;
 	int ret = 0;
 	u32 reg;
+	int wait_clk = 0;
+	int clk_ns = 0;
 
 	/* We don't support transfer larger than the FIFO */
 	if (tfr->len > SUN4I_FIFO_DEPTH)
@@ -261,6 +264,25 @@  static int sun4i_spi_transfer_one(struct spi_master *master,
 
 	sun4i_spi_write(sspi, SUN4I_CLK_CTL_REG, reg);
 
+	/*
+	 * Setup wait time between words.
+	 *
+	 * Wait time is set in SPI_CLK cycles. The SPI hardware needs 3
+	 * additional cycles to setup the wait counter, so the minimum delay
+	 * time is 4 cycles.
+	 */
+	if (spi->word_wait_ns) {
+		clk_ns = DIV_ROUND_UP(1000000000, tfr->speed_hz);
+		wait_clk = DIV_ROUND_UP(spi->word_wait_ns, clk_ns) - 3;
+		if (wait_clk < 1) {
+			wait_clk = 1;
+			dev_dbg(&spi->dev,
+				"using minimum of 4 word wait cycles (%uns)",
+				4 * clk_ns);
+		}
+	}
+	sun4i_spi_write(sspi, SUN4I_WAIT_REG, (u16)wait_clk);
+
 	/* Setup the transfer now... */
 	if (sspi->tx_buf)
 		tx_len = tfr->len;