diff mbox

[v3,2/7] spi: imx: replace fixed timeout with calculated one

Message ID 1446388901-6073-3-git-send-email-anton.bondarenko.sama@gmail.com (mailing list archive)
State New, archived
Headers show

Commit Message

Anton Bondarenko Nov. 1, 2015, 2:41 p.m. UTC
From: Anton Bondarenko <anton_bondarenko@mentor.com>

Fixed timeout value can fire while transaction is ongoing. This may happen
because there are no strict requirements on SPI transaction duration.
Dynamic timeout value is generated based on SCLK and transaction size.

There is also 4 * SCLK delay between TX bursts related to CS change.

Signed-off-by: Anton Bondarenko <anton_bondarenko@mentor.com>
---
 drivers/spi/spi-imx.c | 49 +++++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 41 insertions(+), 8 deletions(-)

Comments

Sascha Hauer Nov. 5, 2015, 8:47 a.m. UTC | #1
On Sun, Nov 01, 2015 at 03:41:36PM +0100, Anton Bondarenko wrote:
> From: Anton Bondarenko <anton_bondarenko@mentor.com>
> 
> Fixed timeout value can fire while transaction is ongoing. This may happen
> because there are no strict requirements on SPI transaction duration.
> Dynamic timeout value is generated based on SCLK and transaction size.
> 
> There is also 4 * SCLK delay between TX bursts related to CS change.
> 
> Signed-off-by: Anton Bondarenko <anton_bondarenko@mentor.com>
> ---
>  drivers/spi/spi-imx.c | 49 +++++++++++++++++++++++++++++++++++++++++--------
>  1 file changed, 41 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/spi/spi-imx.c b/drivers/spi/spi-imx.c
> index bd7b721..9b80c7f 100644
> --- a/drivers/spi/spi-imx.c
> +++ b/drivers/spi/spi-imx.c
> @@ -57,7 +57,6 @@
>  
>  /* The maximum  bytes that a sdma BD can transfer.*/
>  #define MAX_SDMA_BD_BYTES  (1 << 15)
> -#define IMX_DMA_TIMEOUT (msecs_to_jiffies(3000))
>  struct spi_imx_config {
>  	unsigned int speed_hz;
>  	unsigned int bpw;
> @@ -93,6 +92,7 @@ struct spi_imx_data {
>  	struct clk *clk_per;
>  	struct clk *clk_ipg;
>  	unsigned long spi_clk;
> +	unsigned int spi_bus_clk;
>  
>  	unsigned int count;
>  	void (*tx)(struct spi_imx_data *);
> @@ -314,8 +314,7 @@ static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
>  {
>  	u32 ctrl = MX51_ECSPI_CTRL_ENABLE, dma = 0;
>  	u32 cfg = readl(spi_imx->base + MX51_ECSPI_CONFIG);
> -
> -	u32 clk = config->speed_hz, delay;
> +	u32 delay;
>  
>  	/*
>  	 * The hardware seems to have a race condition when changing modes. The
> @@ -327,7 +326,9 @@ static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
>  	ctrl |= MX51_ECSPI_CTRL_MODE_MASK;
>  
>  	/* set clock speed */
> -	ctrl |= mx51_ecspi_clkdiv(spi_imx->spi_clk, config->speed_hz, &clk);
> +	spi_imx->spi_bus_clk = config->speed_hz;
> +	ctrl |= mx51_ecspi_clkdiv(spi_imx->spi_clk, config->speed_hz,
> +				  &spi_imx->spi_bus_clk);
>  
>  	/* set chip select to use */
>  	ctrl |= MX51_ECSPI_CTRL_CS(config->cs);
> @@ -367,7 +368,7 @@ static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
>  	 * the SPI communication as the device on the other end would consider
>  	 * the change of SCLK polarity as a clock tick already.
>  	 */
> -	delay = (2 * 1000000) / clk;
> +	delay = (2 * USEC_PER_SEC) / spi_imx->spi_bus_clk;
>  	if (likely(delay < 10))	/* SCLK is faster than 100 kHz */
>  		udelay(delay);
>  	else			/* SCLK is _very_ slow */
> @@ -890,12 +891,40 @@ static void spi_imx_dma_tx_callback(void *cookie)
>  	complete(&spi_imx->dma_tx_completion);
>  }
>  
> +static int spi_imx_calculate_timeout(struct spi_imx_data *spi_imx, int size)
> +{
> +	unsigned long coef1 = 1;
> +	unsigned long coef2 = MSEC_PER_SEC;
> +	unsigned long timeout = 0;
> +
> +	/* Swap coeficients to avoid div by 0 */
> +	if (spi_imx->spi_bus_clk < MSEC_PER_SEC) {
> +		coef1 = MSEC_PER_SEC;
> +		coef2 = 1;
> +	}
> +
> +	/* Time with actual data transfer */
> +	timeout += DIV_ROUND_UP(8 * size * coef1,
> +				spi_imx->spi_bus_clk / coef2);
> +
> +	/* Take CS change delay related to HW */
> +	timeout += DIV_ROUND_UP((size - 1) * 4 * coef1,
> +				spi_imx->spi_bus_clk / coef2);
> +
> +	/* Add extra second for scheduler related activities */
> +	timeout += MSEC_PER_SEC;
> +
> +	/* Double calculated timeout */
> +	return msecs_to_jiffies(2 * timeout);
> +}

I think you can simplify this function to:

	timeout = size * 8 / spi_imx->spi_bus_clk;
	timeout += 2;

	return msecs_to_jiffies(timeout * MSEC_PER_SEC);

The rationale is that when size * 8 / spi_imx->spi_bus_clk is 0 then we
can add another second and be fine. No need to calculate this more
accurate, in the end it's important to catch the timeout. If we do this
one or two seconds too late doesn't matter.

Sascha
Anton Bondarenko Nov. 10, 2015, 8:20 p.m. UTC | #2
On 05.11.2015 09:47, Sascha Hauer wrote:
>> @@ -890,12 +891,40 @@ static void spi_imx_dma_tx_callback(void *cookie)
>>   	complete(&spi_imx->dma_tx_completion);
>>   }
>>
>> +static int spi_imx_calculate_timeout(struct spi_imx_data *spi_imx, int size)
>> +{
>> +	unsigned long coef1 = 1;
>> +	unsigned long coef2 = MSEC_PER_SEC;
>> +	unsigned long timeout = 0;
>> +
>> +	/* Swap coeficients to avoid div by 0 */
>> +	if (spi_imx->spi_bus_clk < MSEC_PER_SEC) {
>> +		coef1 = MSEC_PER_SEC;
>> +		coef2 = 1;
>> +	}
>> +
>> +	/* Time with actual data transfer */
>> +	timeout += DIV_ROUND_UP(8 * size * coef1,
>> +				spi_imx->spi_bus_clk / coef2);
>> +
>> +	/* Take CS change delay related to HW */
>> +	timeout += DIV_ROUND_UP((size - 1) * 4 * coef1,
>> +				spi_imx->spi_bus_clk / coef2);
>> +
>> +	/* Add extra second for scheduler related activities */
>> +	timeout += MSEC_PER_SEC;
>> +
>> +	/* Double calculated timeout */
>> +	return msecs_to_jiffies(2 * timeout);
>> +}
>
> I think you can simplify this function to:
>
> 	timeout = size * 8 / spi_imx->spi_bus_clk;
> 	timeout += 2;
>
> 	return msecs_to_jiffies(timeout * MSEC_PER_SEC);
>
> The rationale is that when size * 8 / spi_imx->spi_bus_clk is 0 then we
> can add another second and be fine. No need to calculate this more
> accurate, in the end it's important to catch the timeout. If we do this
> one or two seconds too late doesn't matter.
>
> Sascha
>

Sascha,

What about something like this:
	timeout = size * (8 + 4) / spi_imx->spi_bus_clk;
	timeout += 2;

	return msecs_to_jiffies(2 * timeout * MSEC_PER_SEC);

I think we still need to take into account 4 CLKs between each burst.

Regards, Anton
Sascha Hauer Nov. 11, 2015, 8:12 a.m. UTC | #3
On Tue, Nov 10, 2015 at 09:20:06PM +0100, Anton Bondarenko wrote:
> 
> 
> On 05.11.2015 09:47, Sascha Hauer wrote:
> >>@@ -890,12 +891,40 @@ static void spi_imx_dma_tx_callback(void *cookie)
> >>  	complete(&spi_imx->dma_tx_completion);
> >>  }
> >>
> >>+static int spi_imx_calculate_timeout(struct spi_imx_data *spi_imx, int size)
> >>+{
> >>+	unsigned long coef1 = 1;
> >>+	unsigned long coef2 = MSEC_PER_SEC;
> >>+	unsigned long timeout = 0;
> >>+
> >>+	/* Swap coeficients to avoid div by 0 */
> >>+	if (spi_imx->spi_bus_clk < MSEC_PER_SEC) {
> >>+		coef1 = MSEC_PER_SEC;
> >>+		coef2 = 1;
> >>+	}
> >>+
> >>+	/* Time with actual data transfer */
> >>+	timeout += DIV_ROUND_UP(8 * size * coef1,
> >>+				spi_imx->spi_bus_clk / coef2);
> >>+
> >>+	/* Take CS change delay related to HW */
> >>+	timeout += DIV_ROUND_UP((size - 1) * 4 * coef1,
> >>+				spi_imx->spi_bus_clk / coef2);
> >>+
> >>+	/* Add extra second for scheduler related activities */
> >>+	timeout += MSEC_PER_SEC;
> >>+
> >>+	/* Double calculated timeout */
> >>+	return msecs_to_jiffies(2 * timeout);
> >>+}
> >
> >I think you can simplify this function to:
> >
> >	timeout = size * 8 / spi_imx->spi_bus_clk;
> >	timeout += 2;
> >
> >	return msecs_to_jiffies(timeout * MSEC_PER_SEC);
> >
> >The rationale is that when size * 8 / spi_imx->spi_bus_clk is 0 then we
> >can add another second and be fine. No need to calculate this more
> >accurate, in the end it's important to catch the timeout. If we do this
> >one or two seconds too late doesn't matter.
> >
> >Sascha
> >
> 
> Sascha,
> 
> What about something like this:
> 	timeout = size * (8 + 4) / spi_imx->spi_bus_clk;
> 	timeout += 2;
> 
> 	return msecs_to_jiffies(2 * timeout * MSEC_PER_SEC);
> 
> I think we still need to take into account 4 CLKs between each burst.

Fine with me.

Sascha
Anton Bondarenko Nov. 14, 2015, 10:01 a.m. UTC | #4
On 11.11.2015 09:12, Sascha Hauer wrote:
> On Tue, Nov 10, 2015 at 09:20:06PM +0100, Anton Bondarenko wrote:
>>
>>
>> On 05.11.2015 09:47, Sascha Hauer wrote:
>>>> @@ -890,12 +891,40 @@ static void spi_imx_dma_tx_callback(void *cookie)
>>>>   	complete(&spi_imx->dma_tx_completion);
>>>>   }
>>>>
>>>> +static int spi_imx_calculate_timeout(struct spi_imx_data *spi_imx, int size)
>>>> +{
>>>> +	unsigned long coef1 = 1;
>>>> +	unsigned long coef2 = MSEC_PER_SEC;
>>>> +	unsigned long timeout = 0;
>>>> +
>>>> +	/* Swap coeficients to avoid div by 0 */
>>>> +	if (spi_imx->spi_bus_clk < MSEC_PER_SEC) {
>>>> +		coef1 = MSEC_PER_SEC;
>>>> +		coef2 = 1;
>>>> +	}
>>>> +
>>>> +	/* Time with actual data transfer */
>>>> +	timeout += DIV_ROUND_UP(8 * size * coef1,
>>>> +				spi_imx->spi_bus_clk / coef2);
>>>> +
>>>> +	/* Take CS change delay related to HW */
>>>> +	timeout += DIV_ROUND_UP((size - 1) * 4 * coef1,
>>>> +				spi_imx->spi_bus_clk / coef2);
>>>> +
>>>> +	/* Add extra second for scheduler related activities */
>>>> +	timeout += MSEC_PER_SEC;
>>>> +
>>>> +	/* Double calculated timeout */
>>>> +	return msecs_to_jiffies(2 * timeout);
>>>> +}
>>>
>>> I think you can simplify this function to:
>>>
>>> 	timeout = size * 8 / spi_imx->spi_bus_clk;
>>> 	timeout += 2;
>>>
>>> 	return msecs_to_jiffies(timeout * MSEC_PER_SEC);
>>>
>>> The rationale is that when size * 8 / spi_imx->spi_bus_clk is 0 then we
>>> can add another second and be fine. No need to calculate this more
>>> accurate, in the end it's important to catch the timeout. If we do this
>>> one or two seconds too late doesn't matter.
>>>
>>> Sascha
>>>
>>
>> Sascha,
>>
>> What about something like this:
>> 	timeout = size * (8 + 4) / spi_imx->spi_bus_clk;
>> 	timeout += 2;
>>
>> 	return msecs_to_jiffies(2 * timeout * MSEC_PER_SEC);
>>
>> I think we still need to take into account 4 CLKs between each burst.
>
> Fine with me.
>
> Sascha
>
>

Tested new implementation successfully. Will be V4 patchset. Does anyone 
has other comments regarding this commit?

Regards, Anton
diff mbox

Patch

diff --git a/drivers/spi/spi-imx.c b/drivers/spi/spi-imx.c
index bd7b721..9b80c7f 100644
--- a/drivers/spi/spi-imx.c
+++ b/drivers/spi/spi-imx.c
@@ -57,7 +57,6 @@ 
 
 /* The maximum  bytes that a sdma BD can transfer.*/
 #define MAX_SDMA_BD_BYTES  (1 << 15)
-#define IMX_DMA_TIMEOUT (msecs_to_jiffies(3000))
 struct spi_imx_config {
 	unsigned int speed_hz;
 	unsigned int bpw;
@@ -93,6 +92,7 @@  struct spi_imx_data {
 	struct clk *clk_per;
 	struct clk *clk_ipg;
 	unsigned long spi_clk;
+	unsigned int spi_bus_clk;
 
 	unsigned int count;
 	void (*tx)(struct spi_imx_data *);
@@ -314,8 +314,7 @@  static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
 {
 	u32 ctrl = MX51_ECSPI_CTRL_ENABLE, dma = 0;
 	u32 cfg = readl(spi_imx->base + MX51_ECSPI_CONFIG);
-
-	u32 clk = config->speed_hz, delay;
+	u32 delay;
 
 	/*
 	 * The hardware seems to have a race condition when changing modes. The
@@ -327,7 +326,9 @@  static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
 	ctrl |= MX51_ECSPI_CTRL_MODE_MASK;
 
 	/* set clock speed */
-	ctrl |= mx51_ecspi_clkdiv(spi_imx->spi_clk, config->speed_hz, &clk);
+	spi_imx->spi_bus_clk = config->speed_hz;
+	ctrl |= mx51_ecspi_clkdiv(spi_imx->spi_clk, config->speed_hz,
+				  &spi_imx->spi_bus_clk);
 
 	/* set chip select to use */
 	ctrl |= MX51_ECSPI_CTRL_CS(config->cs);
@@ -367,7 +368,7 @@  static int __maybe_unused mx51_ecspi_config(struct spi_imx_data *spi_imx,
 	 * the SPI communication as the device on the other end would consider
 	 * the change of SCLK polarity as a clock tick already.
 	 */
-	delay = (2 * 1000000) / clk;
+	delay = (2 * USEC_PER_SEC) / spi_imx->spi_bus_clk;
 	if (likely(delay < 10))	/* SCLK is faster than 100 kHz */
 		udelay(delay);
 	else			/* SCLK is _very_ slow */
@@ -890,12 +891,40 @@  static void spi_imx_dma_tx_callback(void *cookie)
 	complete(&spi_imx->dma_tx_completion);
 }
 
+static int spi_imx_calculate_timeout(struct spi_imx_data *spi_imx, int size)
+{
+	unsigned long coef1 = 1;
+	unsigned long coef2 = MSEC_PER_SEC;
+	unsigned long timeout = 0;
+
+	/* Swap coeficients to avoid div by 0 */
+	if (spi_imx->spi_bus_clk < MSEC_PER_SEC) {
+		coef1 = MSEC_PER_SEC;
+		coef2 = 1;
+	}
+
+	/* Time with actual data transfer */
+	timeout += DIV_ROUND_UP(8 * size * coef1,
+				spi_imx->spi_bus_clk / coef2);
+
+	/* Take CS change delay related to HW */
+	timeout += DIV_ROUND_UP((size - 1) * 4 * coef1,
+				spi_imx->spi_bus_clk / coef2);
+
+	/* Add extra second for scheduler related activities */
+	timeout += MSEC_PER_SEC;
+
+	/* Double calculated timeout */
+	return msecs_to_jiffies(2 * timeout);
+}
+
 static int spi_imx_dma_transfer(struct spi_imx_data *spi_imx,
 				struct spi_transfer *transfer)
 {
 	struct dma_async_tx_descriptor *desc_tx = NULL, *desc_rx = NULL;
 	int ret;
 	unsigned long timeout;
+	unsigned long transfer_timeout;
 	const int left = transfer->len % spi_imx->wml;
 	struct spi_master *master = spi_imx->bitbang.master;
 	struct sg_table *tx = &transfer->tx_sg, *rx = &transfer->rx_sg;
@@ -956,9 +985,11 @@  static int spi_imx_dma_transfer(struct spi_imx_data *spi_imx,
 	dma_async_issue_pending(master->dma_tx);
 	spi_imx->devtype_data->trigger(spi_imx);
 
+	transfer_timeout = spi_imx_calculate_timeout(spi_imx, transfer->len);
+
 	/* Wait SDMA to finish the data transfer.*/
 	timeout = wait_for_completion_timeout(&spi_imx->dma_tx_completion,
-						IMX_DMA_TIMEOUT);
+						transfer_timeout);
 	if (!timeout) {
 		pr_warn("%s %s: I/O Error in DMA TX\n",
 			dev_driver_string(&master->dev),
@@ -966,8 +997,10 @@  static int spi_imx_dma_transfer(struct spi_imx_data *spi_imx,
 		dmaengine_terminate_all(master->dma_tx);
 		dmaengine_terminate_all(master->dma_rx);
 	} else {
+		transfer_timeout = spi_imx_calculate_timeout(spi_imx,
+							     spi_imx->wml);
 		timeout = wait_for_completion_timeout(
-				&spi_imx->dma_rx_completion, IMX_DMA_TIMEOUT);
+				&spi_imx->dma_rx_completion, transfer_timeout);
 		if (!timeout) {
 			pr_warn("%s %s: I/O Error in DMA RX\n",
 				dev_driver_string(&master->dev),
@@ -989,7 +1022,7 @@  static int spi_imx_dma_transfer(struct spi_imx_data *spi_imx,
 			spi_imx->devtype_data->intctrl(spi_imx, MXC_INT_TCEN);
 
 			timeout = wait_for_completion_timeout(
-					&spi_imx->xfer_done, IMX_DMA_TIMEOUT);
+					&spi_imx->xfer_done, transfer_timeout);
 			if (!timeout) {
 				pr_warn("%s %s: I/O Error in RX tail\n",
 					dev_driver_string(&master->dev),