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
--
To unsubscribe from this list: send the line "unsubscribe linux-spi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
