[PATCH V1 22/26] spi: tegra114: add support for tuning clock delay

[Sowjanya Komatineni]

Tegra SPI controller has TX_CLK_TAP_DELAY and RX_CLK_TAP_DELAY in
COMMAND2 register to tune delay of the clock going out to external
device during transmit and also for the clock coming in from external
device during receive.

TX/RX clock tap delays may vary based on the trace lengths of the
platform design for each of the slaves on the SPI bus.

This patch adds support for configuring TX/RX clock delays specified
through device tree properties.

Signed-off-by: Sowjanya Komatineni <skomatineni@xxxxxxxxxx>
---
 drivers/spi/spi-tegra114.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/drivers/spi/spi-tegra114.c b/drivers/spi/spi-tegra114.c
index e01962344bde..725d60364ec6 100644
--- a/drivers/spi/spi-tegra114.c
+++ b/drivers/spi/spi-tegra114.c
@@ -174,6 +174,8 @@ struct tegra_spi_client_data {
 	int cs_setup_clk_count;
 	int cs_hold_clk_count;
 	int cs_inactive_cycles;
+	int tx_clk_tap_delay;
+	int rx_clk_tap_delay;
 };
 
 struct tegra_spi_data {
@@ -215,8 +217,10 @@ struct tegra_spi_data {
 	u32					command1_reg;
 	u32					dma_control_reg;
 	u32					def_command1_reg;
+	u32					def_command2_reg;
 	u32					spi_cs_timing1;
 	u32					spi_cs_timing2;
+	u8					last_used_cs;
 
 	struct completion			xfer_completion;
 	struct spi_transfer			*curr_xfer;
@@ -780,7 +784,9 @@ static u32 tegra_spi_setup_transfer_one(struct spi_device *spi,
 	u32 speed = t->speed_hz;
 	u8 bits_per_word = t->bits_per_word;
 	u32 command1;
+	u32 command2;
 	int req_mode;
+	u32 tx_tap = 0, rx_tap = 0;
 
 	if (speed != tspi->cur_speed) {
 		clk_set_rate(tspi->clk, speed);
@@ -843,7 +849,18 @@ static u32 tegra_spi_setup_transfer_one(struct spi_device *spi,
 			command1 &= ~(SPI_CS_SW_HW | SPI_CS_SW_VAL);
 		}
 
-		tegra_spi_writel(tspi, 0, SPI_COMMAND2);
+		if (tspi->last_used_cs != spi->chip_select) {
+			if (cdata && cdata->tx_clk_tap_delay)
+				tx_tap = cdata->tx_clk_tap_delay;
+			if (cdata && cdata->rx_clk_tap_delay)
+				rx_tap = cdata->rx_clk_tap_delay;
+			command2 = SPI_TX_TAP_DELAY(tx_tap) |
+				   SPI_RX_TAP_DELAY(rx_tap);
+			if (command2 != tspi->def_command2_reg)
+				tegra_spi_writel(tspi, command2, SPI_COMMAND2);
+			tspi->last_used_cs = spi->chip_select;
+		}
+
 	} else {
 		command1 = tspi->command1_reg;
 		command1 &= ~SPI_BIT_LENGTH(~0);
@@ -923,6 +940,10 @@ static struct tegra_spi_client_data
 			     &cdata->cs_hold_clk_count);
 	of_property_read_u32(slave_np, "nvidia,cs-inactive-cycles",
 			     &cdata->cs_inactive_cycles);
+	of_property_read_u32(slave_np, "nvidia,tx-clk-tap-delay",
+			     &cdata->tx_clk_tap_delay);
+	of_property_read_u32(slave_np, "nvidia,rx-clk-tap-delay",
+			     &cdata->rx_clk_tap_delay);
 	return cdata;
 }
 
@@ -1379,6 +1400,8 @@ static int tegra_spi_probe(struct platform_device *pdev)
 	tegra_spi_writel(tspi, tspi->def_command1_reg, SPI_COMMAND1);
 	tspi->spi_cs_timing1 = tegra_spi_readl(tspi, SPI_CS_TIMING1);
 	tspi->spi_cs_timing2 = tegra_spi_readl(tspi, SPI_CS_TIMING2);
+	tspi->def_command2_reg = tegra_spi_readl(tspi, SPI_COMMAND2);
+	tspi->last_used_cs = master->num_chipselect + 1;
 	pm_runtime_put(&pdev->dev);
 	ret = request_threaded_irq(tspi->irq, tegra_spi_isr,
 				   tegra_spi_isr_thread, IRQF_ONESHOT,
@@ -1451,6 +1474,8 @@ static int tegra_spi_resume(struct device *dev)
 		return ret;
 	}
 	tegra_spi_writel(tspi, tspi->command1_reg, SPI_COMMAND1);
+	tegra_spi_writel(tspi, tspi->def_command2_reg, SPI_COMMAND2);
+	tspi->last_used_cs = master->num_chipselect + 1;
 	pm_runtime_put(dev);
 
 	return spi_master_resume(master);
-- 
2.7.4