spi/tegra114: add spi driver
authorLaxman Dewangan <ldewangan@nvidia.com>
Fri, 22 Feb 2013 12:37:39 +0000 (18:07 +0530)
committerGrant Likely <grant.likely@secretlab.ca>
Sun, 7 Apr 2013 09:08:00 +0000 (10:08 +0100)
Add SPI driver for NVIDIA's Tegra114 SPI controller. This controller
is different than the older SoCs SPI controller in internal design as
well as register interface.

This driver supports the:
- non DMA based transfer for smaller transfer i.e. less than FIFO depth.
- APB DMA based transfer for larger transfer i.e. more than FIFO depth.
- Clock gating through runtime PM callbacks.
- registration through DT only.

Signed-off-by: Laxman Dewangan <ldewangan@nvidia.com>
Reviewed-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Documentation/devicetree/bindings/spi/nvidia,tegra114-spi.txt [new file with mode: 0644]
drivers/spi/Kconfig
drivers/spi/Makefile
drivers/spi/spi-tegra114.c [new file with mode: 0644]

diff --git a/Documentation/devicetree/bindings/spi/nvidia,tegra114-spi.txt b/Documentation/devicetree/bindings/spi/nvidia,tegra114-spi.txt
new file mode 100644 (file)
index 0000000..91ff771
--- /dev/null
@@ -0,0 +1,26 @@
+NVIDIA Tegra114 SPI controller.
+
+Required properties:
+- compatible : should be "nvidia,tegra114-spi".
+- reg: Should contain SPI registers location and length.
+- interrupts: Should contain SPI interrupts.
+- nvidia,dma-request-selector : The Tegra DMA controller's phandle and
+  request selector for this SPI controller.
+- This is also require clock named "spi" as per binding document
+  Documentation/devicetree/bindings/clock/clock-bindings.txt
+
+Recommended properties:
+- spi-max-frequency: Definition as per
+                     Documentation/devicetree/bindings/spi/spi-bus.txt
+Example:
+
+spi@7000d600 {
+       compatible = "nvidia,tegra114-spi";
+       reg = <0x7000d600 0x200>;
+       interrupts = <0 82 0x04>;
+       nvidia,dma-request-selector = <&apbdma 16>;
+       spi-max-frequency = <25000000>;
+       #address-cells = <1>;
+       #size-cells = <0>;
+       status = "disabled";
+};
index 2946ab4..864d87f 100644 (file)
@@ -415,6 +415,14 @@ config SPI_MXS
        help
          SPI driver for Freescale MXS devices.
 
+config SPI_TEGRA114
+       tristate "NVIDIA Tegra114 SPI Controller"
+       depends on ARCH_TEGRA && TEGRA20_APB_DMA
+       help
+         SPI driver for NVIDIA Tegra114 SPI Controller interface. This controller
+         is different than the older SoCs SPI controller and also register interface
+         get changed with this controller.
+
 config SPI_TEGRA20_SFLASH
        tristate "Nvidia Tegra20 Serial flash Controller"
        depends on ARCH_TEGRA
index 604460d..33f9c09 100644 (file)
@@ -65,6 +65,7 @@ obj-$(CONFIG_SPI_SH_HSPI)             += spi-sh-hspi.o
 obj-$(CONFIG_SPI_SH_MSIOF)             += spi-sh-msiof.o
 obj-$(CONFIG_SPI_SH_SCI)               += spi-sh-sci.o
 obj-$(CONFIG_SPI_SIRF)         += spi-sirf.o
+obj-$(CONFIG_SPI_TEGRA114)             += spi-tegra114.o
 obj-$(CONFIG_SPI_TEGRA20_SFLASH)       += spi-tegra20-sflash.o
 obj-$(CONFIG_SPI_TEGRA20_SLINK)                += spi-tegra20-slink.o
 obj-$(CONFIG_SPI_TI_SSP)               += spi-ti-ssp.o
diff --git a/drivers/spi/spi-tegra114.c b/drivers/spi/spi-tegra114.c
new file mode 100644 (file)
index 0000000..598eb45
--- /dev/null
@@ -0,0 +1,1246 @@
+/*
+ * SPI driver for NVIDIA's Tegra114 SPI Controller.
+ *
+ * Copyright (c) 2013, NVIDIA CORPORATION.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/clk.h>
+#include <linux/clk/tegra.h>
+#include <linux/completion.h>
+#include <linux/delay.h>
+#include <linux/dmaengine.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmapool.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/kthread.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/spi/spi.h>
+
+#define SPI_COMMAND1                           0x000
+#define SPI_BIT_LENGTH(x)                      (((x) & 0x1f) << 0)
+#define SPI_PACKED                             (1 << 5)
+#define SPI_TX_EN                              (1 << 11)
+#define SPI_RX_EN                              (1 << 12)
+#define SPI_BOTH_EN_BYTE                       (1 << 13)
+#define SPI_BOTH_EN_BIT                                (1 << 14)
+#define SPI_LSBYTE_FE                          (1 << 15)
+#define SPI_LSBIT_FE                           (1 << 16)
+#define SPI_BIDIROE                            (1 << 17)
+#define SPI_IDLE_SDA_DRIVE_LOW                 (0 << 18)
+#define SPI_IDLE_SDA_DRIVE_HIGH                        (1 << 18)
+#define SPI_IDLE_SDA_PULL_LOW                  (2 << 18)
+#define SPI_IDLE_SDA_PULL_HIGH                 (3 << 18)
+#define SPI_IDLE_SDA_MASK                      (3 << 18)
+#define SPI_CS_SS_VAL                          (1 << 20)
+#define SPI_CS_SW_HW                           (1 << 21)
+/* SPI_CS_POL_INACTIVE bits are default high */
+#define SPI_CS_POL_INACTIVE                    22
+#define SPI_CS_POL_INACTIVE_0                  (1 << 22)
+#define SPI_CS_POL_INACTIVE_1                  (1 << 23)
+#define SPI_CS_POL_INACTIVE_2                  (1 << 24)
+#define SPI_CS_POL_INACTIVE_3                  (1 << 25)
+#define SPI_CS_POL_INACTIVE_MASK               (0xF << 22)
+
+#define SPI_CS_SEL_0                           (0 << 26)
+#define SPI_CS_SEL_1                           (1 << 26)
+#define SPI_CS_SEL_2                           (2 << 26)
+#define SPI_CS_SEL_3                           (3 << 26)
+#define SPI_CS_SEL_MASK                                (3 << 26)
+#define SPI_CS_SEL(x)                          (((x) & 0x3) << 26)
+#define SPI_CONTROL_MODE_0                     (0 << 28)
+#define SPI_CONTROL_MODE_1                     (1 << 28)
+#define SPI_CONTROL_MODE_2                     (2 << 28)
+#define SPI_CONTROL_MODE_3                     (3 << 28)
+#define SPI_CONTROL_MODE_MASK                  (3 << 28)
+#define SPI_MODE_SEL(x)                                (((x) & 0x3) << 28)
+#define SPI_M_S                                        (1 << 30)
+#define SPI_PIO                                        (1 << 31)
+
+#define SPI_COMMAND2                           0x004
+#define SPI_TX_TAP_DELAY(x)                    (((x) & 0x3F) << 6)
+#define SPI_RX_TAP_DELAY(x)                    (((x) & 0x3F) << 0)
+
+#define SPI_CS_TIMING1                         0x008
+#define SPI_SETUP_HOLD(setup, hold)            (((setup) << 4) | (hold))
+#define SPI_CS_SETUP_HOLD(reg, cs, val)                        \
+               ((((val) & 0xFFu) << ((cs) * 8)) |      \
+               ((reg) & ~(0xFFu << ((cs) * 8))))
+
+#define SPI_CS_TIMING2                         0x00C
+#define CYCLES_BETWEEN_PACKETS_0(x)            (((x) & 0x1F) << 0)
+#define CS_ACTIVE_BETWEEN_PACKETS_0            (1 << 5)
+#define CYCLES_BETWEEN_PACKETS_1(x)            (((x) & 0x1F) << 8)
+#define CS_ACTIVE_BETWEEN_PACKETS_1            (1 << 13)
+#define CYCLES_BETWEEN_PACKETS_2(x)            (((x) & 0x1F) << 16)
+#define CS_ACTIVE_BETWEEN_PACKETS_2            (1 << 21)
+#define CYCLES_BETWEEN_PACKETS_3(x)            (((x) & 0x1F) << 24)
+#define CS_ACTIVE_BETWEEN_PACKETS_3            (1 << 29)
+#define SPI_SET_CS_ACTIVE_BETWEEN_PACKETS(reg, cs, val)                \
+               (reg = (((val) & 0x1) << ((cs) * 8 + 5)) |      \
+                       ((reg) & ~(1 << ((cs) * 8 + 5))))
+#define SPI_SET_CYCLES_BETWEEN_PACKETS(reg, cs, val)           \
+               (reg = (((val) & 0xF) << ((cs) * 8)) |          \
+                       ((reg) & ~(0xF << ((cs) * 8))))
+
+#define SPI_TRANS_STATUS                       0x010
+#define SPI_BLK_CNT(val)                       (((val) >> 0) & 0xFFFF)
+#define SPI_SLV_IDLE_COUNT(val)                        (((val) >> 16) & 0xFF)
+#define SPI_RDY                                        (1 << 30)
+
+#define SPI_FIFO_STATUS                                0x014
+#define SPI_RX_FIFO_EMPTY                      (1 << 0)
+#define SPI_RX_FIFO_FULL                       (1 << 1)
+#define SPI_TX_FIFO_EMPTY                      (1 << 2)
+#define SPI_TX_FIFO_FULL                       (1 << 3)
+#define SPI_RX_FIFO_UNF                                (1 << 4)
+#define SPI_RX_FIFO_OVF                                (1 << 5)
+#define SPI_TX_FIFO_UNF                                (1 << 6)
+#define SPI_TX_FIFO_OVF                                (1 << 7)
+#define SPI_ERR                                        (1 << 8)
+#define SPI_TX_FIFO_FLUSH                      (1 << 14)
+#define SPI_RX_FIFO_FLUSH                      (1 << 15)
+#define SPI_TX_FIFO_EMPTY_COUNT(val)           (((val) >> 16) & 0x7F)
+#define SPI_RX_FIFO_FULL_COUNT(val)            (((val) >> 23) & 0x7F)
+#define SPI_FRAME_END                          (1 << 30)
+#define SPI_CS_INACTIVE                                (1 << 31)
+
+#define SPI_FIFO_ERROR                         (SPI_RX_FIFO_UNF | \
+                       SPI_RX_FIFO_OVF | SPI_TX_FIFO_UNF | SPI_TX_FIFO_OVF)
+#define SPI_FIFO_EMPTY                 (SPI_RX_FIFO_EMPTY | SPI_TX_FIFO_EMPTY)
+
+#define SPI_TX_DATA                            0x018
+#define SPI_RX_DATA                            0x01C
+
+#define SPI_DMA_CTL                            0x020
+#define SPI_TX_TRIG_1                          (0 << 15)
+#define SPI_TX_TRIG_4                          (1 << 15)
+#define SPI_TX_TRIG_8                          (2 << 15)
+#define SPI_TX_TRIG_16                         (3 << 15)
+#define SPI_TX_TRIG_MASK                       (3 << 15)
+#define SPI_RX_TRIG_1                          (0 << 19)
+#define SPI_RX_TRIG_4                          (1 << 19)
+#define SPI_RX_TRIG_8                          (2 << 19)
+#define SPI_RX_TRIG_16                         (3 << 19)
+#define SPI_RX_TRIG_MASK                       (3 << 19)
+#define SPI_IE_TX                              (1 << 28)
+#define SPI_IE_RX                              (1 << 29)
+#define SPI_CONT                               (1 << 30)
+#define SPI_DMA                                        (1 << 31)
+#define SPI_DMA_EN                             SPI_DMA
+
+#define SPI_DMA_BLK                            0x024
+#define SPI_DMA_BLK_SET(x)                     (((x) & 0xFFFF) << 0)
+
+#define SPI_TX_FIFO                            0x108
+#define SPI_RX_FIFO                            0x188
+#define MAX_CHIP_SELECT                                4
+#define SPI_FIFO_DEPTH                         64
+#define DATA_DIR_TX                            (1 << 0)
+#define DATA_DIR_RX                            (1 << 1)
+
+#define SPI_DMA_TIMEOUT                                (msecs_to_jiffies(1000))
+#define DEFAULT_SPI_DMA_BUF_LEN                        (16*1024)
+#define TX_FIFO_EMPTY_COUNT_MAX                        SPI_TX_FIFO_EMPTY_COUNT(0x40)
+#define RX_FIFO_FULL_COUNT_ZERO                        SPI_RX_FIFO_FULL_COUNT(0)
+#define MAX_HOLD_CYCLES                                16
+#define SPI_DEFAULT_SPEED                      25000000
+
+#define MAX_CHIP_SELECT                                4
+#define SPI_FIFO_DEPTH                         64
+
+struct tegra_spi_data {
+       struct device                           *dev;
+       struct spi_master                       *master;
+       spinlock_t                              lock;
+
+       struct clk                              *clk;
+       void __iomem                            *base;
+       phys_addr_t                             phys;
+       unsigned                                irq;
+       int                                     dma_req_sel;
+       u32                                     spi_max_frequency;
+       u32                                     cur_speed;
+
+       struct spi_device                       *cur_spi;
+       unsigned                                cur_pos;
+       unsigned                                cur_len;
+       unsigned                                words_per_32bit;
+       unsigned                                bytes_per_word;
+       unsigned                                curr_dma_words;
+       unsigned                                cur_direction;
+
+       unsigned                                cur_rx_pos;
+       unsigned                                cur_tx_pos;
+
+       unsigned                                dma_buf_size;
+       unsigned                                max_buf_size;
+       bool                                    is_curr_dma_xfer;
+
+       struct completion                       rx_dma_complete;
+       struct completion                       tx_dma_complete;
+
+       u32                                     tx_status;
+       u32                                     rx_status;
+       u32                                     status_reg;
+       bool                                    is_packed;
+       unsigned long                           packed_size;
+
+       u32                                     command1_reg;
+       u32                                     dma_control_reg;
+       u32                                     def_command1_reg;
+       u32                                     spi_cs_timing;
+
+       struct completion                       xfer_completion;
+       struct spi_transfer                     *curr_xfer;
+       struct dma_chan                         *rx_dma_chan;
+       u32                                     *rx_dma_buf;
+       dma_addr_t                              rx_dma_phys;
+       struct dma_async_tx_descriptor          *rx_dma_desc;
+
+       struct dma_chan                         *tx_dma_chan;
+       u32                                     *tx_dma_buf;
+       dma_addr_t                              tx_dma_phys;
+       struct dma_async_tx_descriptor          *tx_dma_desc;
+};
+
+static int tegra_spi_runtime_suspend(struct device *dev);
+static int tegra_spi_runtime_resume(struct device *dev);
+
+static inline unsigned long tegra_spi_readl(struct tegra_spi_data *tspi,
+               unsigned long reg)
+{
+       return readl(tspi->base + reg);
+}
+
+static inline void tegra_spi_writel(struct tegra_spi_data *tspi,
+               unsigned long val, unsigned long reg)
+{
+       writel(val, tspi->base + reg);
+
+       /* Read back register to make sure that register writes completed */
+       if (reg != SPI_TX_FIFO)
+               readl(tspi->base + SPI_COMMAND1);
+}
+
+static void tegra_spi_clear_status(struct tegra_spi_data *tspi)
+{
+       unsigned long val;
+
+       /* Write 1 to clear status register */
+       val = tegra_spi_readl(tspi, SPI_TRANS_STATUS);
+       tegra_spi_writel(tspi, val, SPI_TRANS_STATUS);
+
+       /* Clear fifo status error if any */
+       val = tegra_spi_readl(tspi, SPI_FIFO_STATUS);
+       if (val & SPI_ERR)
+               tegra_spi_writel(tspi, SPI_ERR | SPI_FIFO_ERROR,
+                               SPI_FIFO_STATUS);
+}
+
+static unsigned tegra_spi_calculate_curr_xfer_param(
+       struct spi_device *spi, struct tegra_spi_data *tspi,
+       struct spi_transfer *t)
+{
+       unsigned remain_len = t->len - tspi->cur_pos;
+       unsigned max_word;
+       unsigned bits_per_word = t->bits_per_word;
+       unsigned max_len;
+       unsigned total_fifo_words;
+
+       tspi->bytes_per_word = (bits_per_word - 1) / 8 + 1;
+
+       if (bits_per_word == 8 || bits_per_word == 16) {
+               tspi->is_packed = 1;
+               tspi->words_per_32bit = 32/bits_per_word;
+       } else {
+               tspi->is_packed = 0;
+               tspi->words_per_32bit = 1;
+       }
+
+       if (tspi->is_packed) {
+               max_len = min(remain_len, tspi->max_buf_size);
+               tspi->curr_dma_words = max_len/tspi->bytes_per_word;
+               total_fifo_words = (max_len + 3) / 4;
+       } else {
+               max_word = (remain_len - 1) / tspi->bytes_per_word + 1;
+               max_word = min(max_word, tspi->max_buf_size/4);
+               tspi->curr_dma_words = max_word;
+               total_fifo_words = max_word;
+       }
+       return total_fifo_words;
+}
+
+static unsigned tegra_spi_fill_tx_fifo_from_client_txbuf(
+       struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned nbytes;
+       unsigned tx_empty_count;
+       unsigned long fifo_status;
+       unsigned max_n_32bit;
+       unsigned i, count;
+       unsigned long x;
+       unsigned int written_words;
+       unsigned fifo_words_left;
+       u8 *tx_buf = (u8 *)t->tx_buf + tspi->cur_tx_pos;
+
+       fifo_status = tegra_spi_readl(tspi, SPI_FIFO_STATUS);
+       tx_empty_count = SPI_TX_FIFO_EMPTY_COUNT(fifo_status);
+
+       if (tspi->is_packed) {
+               fifo_words_left = tx_empty_count * tspi->words_per_32bit;
+               written_words = min(fifo_words_left, tspi->curr_dma_words);
+               nbytes = written_words * tspi->bytes_per_word;
+               max_n_32bit = DIV_ROUND_UP(nbytes, 4);
+               for (count = 0; count < max_n_32bit; count++) {
+                       x = 0;
+                       for (i = 0; (i < 4) && nbytes; i++, nbytes--)
+                               x |= (*tx_buf++) << (i*8);
+                       tegra_spi_writel(tspi, x, SPI_TX_FIFO);
+               }
+       } else {
+               max_n_32bit = min(tspi->curr_dma_words,  tx_empty_count);
+               written_words = max_n_32bit;
+               nbytes = written_words * tspi->bytes_per_word;
+               for (count = 0; count < max_n_32bit; count++) {
+                       x = 0;
+                       for (i = 0; nbytes && (i < tspi->bytes_per_word);
+                                                       i++, nbytes--)
+                               x |= ((*tx_buf++) << i*8);
+                       tegra_spi_writel(tspi, x, SPI_TX_FIFO);
+               }
+       }
+       tspi->cur_tx_pos += written_words * tspi->bytes_per_word;
+       return written_words;
+}
+
+static unsigned int tegra_spi_read_rx_fifo_to_client_rxbuf(
+               struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned rx_full_count;
+       unsigned long fifo_status;
+       unsigned i, count;
+       unsigned long x;
+       unsigned int read_words = 0;
+       unsigned len;
+       u8 *rx_buf = (u8 *)t->rx_buf + tspi->cur_rx_pos;
+
+       fifo_status = tegra_spi_readl(tspi, SPI_FIFO_STATUS);
+       rx_full_count = SPI_RX_FIFO_FULL_COUNT(fifo_status);
+       if (tspi->is_packed) {
+               len = tspi->curr_dma_words * tspi->bytes_per_word;
+               for (count = 0; count < rx_full_count; count++) {
+                       x = tegra_spi_readl(tspi, SPI_RX_FIFO);
+                       for (i = 0; len && (i < 4); i++, len--)
+                               *rx_buf++ = (x >> i*8) & 0xFF;
+               }
+               tspi->cur_rx_pos += tspi->curr_dma_words * tspi->bytes_per_word;
+               read_words += tspi->curr_dma_words;
+       } else {
+               unsigned int rx_mask;
+               unsigned int bits_per_word = t->bits_per_word;
+
+               rx_mask = (1 << bits_per_word) - 1;
+               for (count = 0; count < rx_full_count; count++) {
+                       x = tegra_spi_readl(tspi, SPI_RX_FIFO);
+                       x &= rx_mask;
+                       for (i = 0; (i < tspi->bytes_per_word); i++)
+                               *rx_buf++ = (x >> (i*8)) & 0xFF;
+               }
+               tspi->cur_rx_pos += rx_full_count * tspi->bytes_per_word;
+               read_words += rx_full_count;
+       }
+       return read_words;
+}
+
+static void tegra_spi_copy_client_txbuf_to_spi_txbuf(
+               struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned len;
+
+       /* Make the dma buffer to read by cpu */
+       dma_sync_single_for_cpu(tspi->dev, tspi->tx_dma_phys,
+                               tspi->dma_buf_size, DMA_TO_DEVICE);
+
+       if (tspi->is_packed) {
+               len = tspi->curr_dma_words * tspi->bytes_per_word;
+               memcpy(tspi->tx_dma_buf, t->tx_buf + tspi->cur_pos, len);
+       } else {
+               unsigned int i;
+               unsigned int count;
+               u8 *tx_buf = (u8 *)t->tx_buf + tspi->cur_tx_pos;
+               unsigned consume = tspi->curr_dma_words * tspi->bytes_per_word;
+               unsigned int x;
+
+               for (count = 0; count < tspi->curr_dma_words; count++) {
+                       x = 0;
+                       for (i = 0; consume && (i < tspi->bytes_per_word);
+                                                       i++, consume--)
+                               x |= ((*tx_buf++) << i * 8);
+                       tspi->tx_dma_buf[count] = x;
+               }
+       }
+       tspi->cur_tx_pos += tspi->curr_dma_words * tspi->bytes_per_word;
+
+       /* Make the dma buffer to read by dma */
+       dma_sync_single_for_device(tspi->dev, tspi->tx_dma_phys,
+                               tspi->dma_buf_size, DMA_TO_DEVICE);
+}
+
+static void tegra_spi_copy_spi_rxbuf_to_client_rxbuf(
+               struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned len;
+
+       /* Make the dma buffer to read by cpu */
+       dma_sync_single_for_cpu(tspi->dev, tspi->rx_dma_phys,
+               tspi->dma_buf_size, DMA_FROM_DEVICE);
+
+       if (tspi->is_packed) {
+               len = tspi->curr_dma_words * tspi->bytes_per_word;
+               memcpy(t->rx_buf + tspi->cur_rx_pos, tspi->rx_dma_buf, len);
+       } else {
+               unsigned int i;
+               unsigned int count;
+               unsigned char *rx_buf = t->rx_buf + tspi->cur_rx_pos;
+               unsigned int x;
+               unsigned int rx_mask;
+               unsigned int bits_per_word = t->bits_per_word;
+
+               rx_mask = (1 << bits_per_word) - 1;
+               for (count = 0; count < tspi->curr_dma_words; count++) {
+                       x = tspi->rx_dma_buf[count];
+                       x &= rx_mask;
+                       for (i = 0; (i < tspi->bytes_per_word); i++)
+                               *rx_buf++ = (x >> (i*8)) & 0xFF;
+               }
+       }
+       tspi->cur_rx_pos += tspi->curr_dma_words * tspi->bytes_per_word;
+
+       /* Make the dma buffer to read by dma */
+       dma_sync_single_for_device(tspi->dev, tspi->rx_dma_phys,
+               tspi->dma_buf_size, DMA_FROM_DEVICE);
+}
+
+static void tegra_spi_dma_complete(void *args)
+{
+       struct completion *dma_complete = args;
+
+       complete(dma_complete);
+}
+
+static int tegra_spi_start_tx_dma(struct tegra_spi_data *tspi, int len)
+{
+       INIT_COMPLETION(tspi->tx_dma_complete);
+       tspi->tx_dma_desc = dmaengine_prep_slave_single(tspi->tx_dma_chan,
+                               tspi->tx_dma_phys, len, DMA_MEM_TO_DEV,
+                               DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
+       if (!tspi->tx_dma_desc) {
+               dev_err(tspi->dev, "Not able to get desc for Tx\n");
+               return -EIO;
+       }
+
+       tspi->tx_dma_desc->callback = tegra_spi_dma_complete;
+       tspi->tx_dma_desc->callback_param = &tspi->tx_dma_complete;
+
+       dmaengine_submit(tspi->tx_dma_desc);
+       dma_async_issue_pending(tspi->tx_dma_chan);
+       return 0;
+}
+
+static int tegra_spi_start_rx_dma(struct tegra_spi_data *tspi, int len)
+{
+       INIT_COMPLETION(tspi->rx_dma_complete);
+       tspi->rx_dma_desc = dmaengine_prep_slave_single(tspi->rx_dma_chan,
+                               tspi->rx_dma_phys, len, DMA_DEV_TO_MEM,
+                               DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
+       if (!tspi->rx_dma_desc) {
+               dev_err(tspi->dev, "Not able to get desc for Rx\n");
+               return -EIO;
+       }
+
+       tspi->rx_dma_desc->callback = tegra_spi_dma_complete;
+       tspi->rx_dma_desc->callback_param = &tspi->rx_dma_complete;
+
+       dmaengine_submit(tspi->rx_dma_desc);
+       dma_async_issue_pending(tspi->rx_dma_chan);
+       return 0;
+}
+
+static int tegra_spi_start_dma_based_transfer(
+               struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned long val;
+       unsigned int len;
+       int ret = 0;
+       unsigned long status;
+
+       /* Make sure that Rx and Tx fifo are empty */
+       status = tegra_spi_readl(tspi, SPI_FIFO_STATUS);
+       if ((status & SPI_FIFO_EMPTY) != SPI_FIFO_EMPTY) {
+               dev_err(tspi->dev,
+                       "Rx/Tx fifo are not empty status 0x%08lx\n", status);
+               return -EIO;
+       }
+
+       val = SPI_DMA_BLK_SET(tspi->curr_dma_words - 1);
+       tegra_spi_writel(tspi, val, SPI_DMA_BLK);
+
+       if (tspi->is_packed)
+               len = DIV_ROUND_UP(tspi->curr_dma_words * tspi->bytes_per_word,
+                                       4) * 4;
+       else
+               len = tspi->curr_dma_words * 4;
+
+       /* Set attention level based on length of transfer */
+       if (len & 0xF)
+               val |= SPI_TX_TRIG_1 | SPI_RX_TRIG_1;
+       else if (((len) >> 4) & 0x1)
+               val |= SPI_TX_TRIG_4 | SPI_RX_TRIG_4;
+       else
+               val |= SPI_TX_TRIG_8 | SPI_RX_TRIG_8;
+
+       if (tspi->cur_direction & DATA_DIR_TX)
+               val |= SPI_IE_TX;
+
+       if (tspi->cur_direction & DATA_DIR_RX)
+               val |= SPI_IE_RX;
+
+       tegra_spi_writel(tspi, val, SPI_DMA_CTL);
+       tspi->dma_control_reg = val;
+
+       if (tspi->cur_direction & DATA_DIR_TX) {
+               tegra_spi_copy_client_txbuf_to_spi_txbuf(tspi, t);
+               ret = tegra_spi_start_tx_dma(tspi, len);
+               if (ret < 0) {
+                       dev_err(tspi->dev,
+                               "Starting tx dma failed, err %d\n", ret);
+                       return ret;
+               }
+       }
+
+       if (tspi->cur_direction & DATA_DIR_RX) {
+               /* Make the dma buffer to read by dma */
+               dma_sync_single_for_device(tspi->dev, tspi->rx_dma_phys,
+                               tspi->dma_buf_size, DMA_FROM_DEVICE);
+
+               ret = tegra_spi_start_rx_dma(tspi, len);
+               if (ret < 0) {
+                       dev_err(tspi->dev,
+                               "Starting rx dma failed, err %d\n", ret);
+                       if (tspi->cur_direction & DATA_DIR_TX)
+                               dmaengine_terminate_all(tspi->tx_dma_chan);
+                       return ret;
+               }
+       }
+       tspi->is_curr_dma_xfer = true;
+       tspi->dma_control_reg = val;
+
+       val |= SPI_DMA_EN;
+       tegra_spi_writel(tspi, val, SPI_DMA_CTL);
+       return ret;
+}
+
+static int tegra_spi_start_cpu_based_transfer(
+               struct tegra_spi_data *tspi, struct spi_transfer *t)
+{
+       unsigned long val;
+       unsigned cur_words;
+
+       if (tspi->cur_direction & DATA_DIR_TX)
+               cur_words = tegra_spi_fill_tx_fifo_from_client_txbuf(tspi, t);
+       else
+               cur_words = tspi->curr_dma_words;
+
+       val = SPI_DMA_BLK_SET(cur_words - 1);
+       tegra_spi_writel(tspi, val, SPI_DMA_BLK);
+
+       val = 0;
+       if (tspi->cur_direction & DATA_DIR_TX)
+               val |= SPI_IE_TX;
+
+       if (tspi->cur_direction & DATA_DIR_RX)
+               val |= SPI_IE_RX;
+
+       tegra_spi_writel(tspi, val, SPI_DMA_CTL);
+       tspi->dma_control_reg = val;
+
+       tspi->is_curr_dma_xfer = false;
+
+       val |= SPI_DMA_EN;
+       tegra_spi_writel(tspi, val, SPI_DMA_CTL);
+       return 0;
+}
+
+static int tegra_spi_init_dma_param(struct tegra_spi_data *tspi,
+                       bool dma_to_memory)
+{
+       struct dma_chan *dma_chan;
+       u32 *dma_buf;
+       dma_addr_t dma_phys;
+       int ret;
+       struct dma_slave_config dma_sconfig;
+       dma_cap_mask_t mask;
+
+       dma_cap_zero(mask);
+       dma_cap_set(DMA_SLAVE, mask);
+       dma_chan = dma_request_channel(mask, NULL, NULL);
+       if (!dma_chan) {
+               dev_err(tspi->dev,
+                       "Dma channel is not available, will try later\n");
+               return -EPROBE_DEFER;
+       }
+
+       dma_buf = dma_alloc_coherent(tspi->dev, tspi->dma_buf_size,
+                               &dma_phys, GFP_KERNEL);
+       if (!dma_buf) {
+               dev_err(tspi->dev, " Not able to allocate the dma buffer\n");
+               dma_release_channel(dma_chan);
+               return -ENOMEM;
+       }
+
+       dma_sconfig.slave_id = tspi->dma_req_sel;
+       if (dma_to_memory) {
+               dma_sconfig.src_addr = tspi->phys + SPI_RX_FIFO;
+               dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+               dma_sconfig.src_maxburst = 0;
+       } else {
+               dma_sconfig.dst_addr = tspi->phys + SPI_TX_FIFO;
+               dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+               dma_sconfig.dst_maxburst = 0;
+       }
+
+       ret = dmaengine_slave_config(dma_chan, &dma_sconfig);
+       if (ret)
+               goto scrub;
+       if (dma_to_memory) {
+               tspi->rx_dma_chan = dma_chan;
+               tspi->rx_dma_buf = dma_buf;
+               tspi->rx_dma_phys = dma_phys;
+       } else {
+               tspi->tx_dma_chan = dma_chan;
+               tspi->tx_dma_buf = dma_buf;
+               tspi->tx_dma_phys = dma_phys;
+       }
+       return 0;
+
+scrub:
+       dma_free_coherent(tspi->dev, tspi->dma_buf_size, dma_buf, dma_phys);
+       dma_release_channel(dma_chan);
+       return ret;
+}
+
+static void tegra_spi_deinit_dma_param(struct tegra_spi_data *tspi,
+       bool dma_to_memory)
+{
+       u32 *dma_buf;
+       dma_addr_t dma_phys;
+       struct dma_chan *dma_chan;
+
+       if (dma_to_memory) {
+               dma_buf = tspi->rx_dma_buf;
+               dma_chan = tspi->rx_dma_chan;
+               dma_phys = tspi->rx_dma_phys;
+               tspi->rx_dma_chan = NULL;
+               tspi->rx_dma_buf = NULL;
+       } else {
+               dma_buf = tspi->tx_dma_buf;
+               dma_chan = tspi->tx_dma_chan;
+               dma_phys = tspi->tx_dma_phys;
+               tspi->tx_dma_buf = NULL;
+               tspi->tx_dma_chan = NULL;
+       }
+       if (!dma_chan)
+               return;
+
+       dma_free_coherent(tspi->dev, tspi->dma_buf_size, dma_buf, dma_phys);
+       dma_release_channel(dma_chan);
+}
+
+static int tegra_spi_start_transfer_one(struct spi_device *spi,
+               struct spi_transfer *t, bool is_first_of_msg,
+               bool is_single_xfer)
+{
+       struct tegra_spi_data *tspi = spi_master_get_devdata(spi->master);
+       u32 speed = t->speed_hz;
+       u8 bits_per_word = t->bits_per_word;
+       unsigned total_fifo_words;
+       int ret;
+       unsigned long command1;
+       int req_mode;
+
+       if (speed != tspi->cur_speed) {
+               clk_set_rate(tspi->clk, speed);
+               tspi->cur_speed = speed;
+       }
+
+       tspi->cur_spi = spi;
+       tspi->cur_pos = 0;
+       tspi->cur_rx_pos = 0;
+       tspi->cur_tx_pos = 0;
+       tspi->curr_xfer = t;
+       total_fifo_words = tegra_spi_calculate_curr_xfer_param(spi, tspi, t);
+
+       if (is_first_of_msg) {
+               tegra_spi_clear_status(tspi);
+
+               command1 = tspi->def_command1_reg;
+               command1 |= SPI_BIT_LENGTH(bits_per_word - 1);
+
+               command1 &= ~SPI_CONTROL_MODE_MASK;
+               req_mode = spi->mode & 0x3;
+               if (req_mode == SPI_MODE_0)
+                       command1 |= SPI_CONTROL_MODE_0;
+               else if (req_mode == SPI_MODE_1)
+                       command1 |= SPI_CONTROL_MODE_1;
+               else if (req_mode == SPI_MODE_2)
+                       command1 |= SPI_CONTROL_MODE_2;
+               else if (req_mode == SPI_MODE_3)
+                       command1 |= SPI_CONTROL_MODE_3;
+
+               tegra_spi_writel(tspi, command1, SPI_COMMAND1);
+
+               command1 |= SPI_CS_SW_HW;
+               if (spi->mode & SPI_CS_HIGH)
+                       command1 |= SPI_CS_SS_VAL;
+               else
+                       command1 &= ~SPI_CS_SS_VAL;
+
+               tegra_spi_writel(tspi, 0, SPI_COMMAND2);
+       } else {
+               command1 = tspi->command1_reg;
+               command1 &= ~SPI_BIT_LENGTH(~0);
+               command1 |= SPI_BIT_LENGTH(bits_per_word - 1);
+       }
+
+       if (tspi->is_packed)
+               command1 |= SPI_PACKED;
+
+       command1 &= ~(SPI_CS_SEL_MASK | SPI_TX_EN | SPI_RX_EN);
+       tspi->cur_direction = 0;
+       if (t->rx_buf) {
+               command1 |= SPI_RX_EN;
+               tspi->cur_direction |= DATA_DIR_RX;
+       }
+       if (t->tx_buf) {
+               command1 |= SPI_TX_EN;
+               tspi->cur_direction |= DATA_DIR_TX;
+       }
+       command1 |= SPI_CS_SEL(spi->chip_select);
+       tegra_spi_writel(tspi, command1, SPI_COMMAND1);
+       tspi->command1_reg = command1;
+
+       dev_dbg(tspi->dev, "The def 0x%x and written 0x%lx\n",
+                               tspi->def_command1_reg, command1);
+
+       if (total_fifo_words > SPI_FIFO_DEPTH)
+               ret = tegra_spi_start_dma_based_transfer(tspi, t);
+       else
+               ret = tegra_spi_start_cpu_based_transfer(tspi, t);
+       return ret;
+}
+
+static int tegra_spi_setup(struct spi_device *spi)
+{
+       struct tegra_spi_data *tspi = spi_master_get_devdata(spi->master);
+       unsigned long val;
+       unsigned long flags;
+       int ret;
+       unsigned int cs_pol_bit[MAX_CHIP_SELECT] = {
+                       SPI_CS_POL_INACTIVE_0,
+                       SPI_CS_POL_INACTIVE_1,
+                       SPI_CS_POL_INACTIVE_2,
+                       SPI_CS_POL_INACTIVE_3,
+       };
+
+       dev_dbg(&spi->dev, "setup %d bpw, %scpol, %scpha, %dHz\n",
+               spi->bits_per_word,
+               spi->mode & SPI_CPOL ? "" : "~",
+               spi->mode & SPI_CPHA ? "" : "~",
+               spi->max_speed_hz);
+
+       BUG_ON(spi->chip_select >= MAX_CHIP_SELECT);
+
+       /* Set speed to the spi max fequency if spi device has not set */
+       spi->max_speed_hz = spi->max_speed_hz ? : tspi->spi_max_frequency;
+
+       ret = pm_runtime_get_sync(tspi->dev);
+       if (ret < 0) {
+               dev_err(tspi->dev, "pm runtime failed, e = %d\n", ret);
+               return ret;
+       }
+
+       spin_lock_irqsave(&tspi->lock, flags);
+       val = tspi->def_command1_reg;
+       if (spi->mode & SPI_CS_HIGH)
+               val &= ~cs_pol_bit[spi->chip_select];
+       else
+               val |= cs_pol_bit[spi->chip_select];
+       tspi->def_command1_reg = val;
+       tegra_spi_writel(tspi, tspi->def_command1_reg, SPI_COMMAND1);
+       spin_unlock_irqrestore(&tspi->lock, flags);
+
+       pm_runtime_put(tspi->dev);
+       return 0;
+}
+
+static int tegra_spi_transfer_one_message(struct spi_master *master,
+                       struct spi_message *msg)
+{
+       bool is_first_msg = true;
+       int single_xfer;
+       struct tegra_spi_data *tspi = spi_master_get_devdata(master);
+       struct spi_transfer *xfer;
+       struct spi_device *spi = msg->spi;
+       int ret;
+
+       msg->status = 0;
+       msg->actual_length = 0;
+
+       ret = pm_runtime_get_sync(tspi->dev);
+       if (ret < 0) {
+               dev_err(tspi->dev, "runtime PM get failed: %d\n", ret);
+               msg->status = ret;
+               spi_finalize_current_message(master);
+               return ret;
+       }
+
+       single_xfer = list_is_singular(&msg->transfers);
+       list_for_each_entry(xfer, &msg->transfers, transfer_list) {
+               INIT_COMPLETION(tspi->xfer_completion);
+               ret = tegra_spi_start_transfer_one(spi, xfer,
+                                       is_first_msg, single_xfer);
+               if (ret < 0) {
+                       dev_err(tspi->dev,
+                               "spi can not start transfer, err %d\n", ret);
+                       goto exit;
+               }
+               is_first_msg = false;
+               ret = wait_for_completion_timeout(&tspi->xfer_completion,
+                                               SPI_DMA_TIMEOUT);
+               if (WARN_ON(ret == 0)) {
+                       dev_err(tspi->dev,
+                               "spi trasfer timeout, err %d\n", ret);
+                       ret = -EIO;
+                       goto exit;
+               }
+
+               if (tspi->tx_status ||  tspi->rx_status) {
+                       dev_err(tspi->dev, "Error in Transfer\n");
+                       ret = -EIO;
+                       goto exit;
+               }
+               msg->actual_length += xfer->len;
+               if (xfer->cs_change && xfer->delay_usecs) {
+                       tegra_spi_writel(tspi, tspi->def_command1_reg,
+                                       SPI_COMMAND1);
+                       udelay(xfer->delay_usecs);
+               }
+       }
+       ret = 0;
+exit:
+       tegra_spi_writel(tspi, tspi->def_command1_reg, SPI_COMMAND1);
+       pm_runtime_put(tspi->dev);
+       msg->status = ret;
+       spi_finalize_current_message(master);
+       return ret;
+}
+
+static irqreturn_t handle_cpu_based_xfer(struct tegra_spi_data *tspi)
+{
+       struct spi_transfer *t = tspi->curr_xfer;
+       unsigned long flags;
+
+       spin_lock_irqsave(&tspi->lock, flags);
+       if (tspi->tx_status ||  tspi->rx_status) {
+               dev_err(tspi->dev, "CpuXfer ERROR bit set 0x%x\n",
+                       tspi->status_reg);
+               dev_err(tspi->dev, "CpuXfer 0x%08x:0x%08x\n",
+                       tspi->command1_reg, tspi->dma_control_reg);
+               tegra_periph_reset_assert(tspi->clk);
+               udelay(2);
+               tegra_periph_reset_deassert(tspi->clk);
+               complete(&tspi->xfer_completion);
+               goto exit;
+       }
+
+       if (tspi->cur_direction & DATA_DIR_RX)
+               tegra_spi_read_rx_fifo_to_client_rxbuf(tspi, t);
+
+       if (tspi->cur_direction & DATA_DIR_TX)
+               tspi->cur_pos = tspi->cur_tx_pos;
+       else
+               tspi->cur_pos = tspi->cur_rx_pos;
+
+       if (tspi->cur_pos == t->len) {
+               complete(&tspi->xfer_completion);
+               goto exit;
+       }
+
+       tegra_spi_calculate_curr_xfer_param(tspi->cur_spi, tspi, t);
+       tegra_spi_start_cpu_based_transfer(tspi, t);
+exit:
+       spin_unlock_irqrestore(&tspi->lock, flags);
+       return IRQ_HANDLED;
+}
+
+static irqreturn_t handle_dma_based_xfer(struct tegra_spi_data *tspi)
+{
+       struct spi_transfer *t = tspi->curr_xfer;
+       long wait_status;
+       int err = 0;
+       unsigned total_fifo_words;
+       unsigned long flags;
+
+       /* Abort dmas if any error */
+       if (tspi->cur_direction & DATA_DIR_TX) {
+               if (tspi->tx_status) {
+                       dmaengine_terminate_all(tspi->tx_dma_chan);
+                       err += 1;
+               } else {
+                       wait_status = wait_for_completion_interruptible_timeout(
+                               &tspi->tx_dma_complete, SPI_DMA_TIMEOUT);
+                       if (wait_status <= 0) {
+                               dmaengine_terminate_all(tspi->tx_dma_chan);
+                               dev_err(tspi->dev, "TxDma Xfer failed\n");
+                               err += 1;
+                       }
+               }
+       }
+
+       if (tspi->cur_direction & DATA_DIR_RX) {
+               if (tspi->rx_status) {
+                       dmaengine_terminate_all(tspi->rx_dma_chan);
+                       err += 2;
+               } else {
+                       wait_status = wait_for_completion_interruptible_timeout(
+                               &tspi->rx_dma_complete, SPI_DMA_TIMEOUT);
+                       if (wait_status <= 0) {
+                               dmaengine_terminate_all(tspi->rx_dma_chan);
+                               dev_err(tspi->dev, "RxDma Xfer failed\n");
+                               err += 2;
+                       }
+               }
+       }
+
+       spin_lock_irqsave(&tspi->lock, flags);
+       if (err) {
+               dev_err(tspi->dev, "DmaXfer: ERROR bit set 0x%x\n",
+                       tspi->status_reg);
+               dev_err(tspi->dev, "DmaXfer 0x%08x:0x%08x\n",
+                       tspi->command1_reg, tspi->dma_control_reg);
+               tegra_periph_reset_assert(tspi->clk);
+               udelay(2);
+               tegra_periph_reset_deassert(tspi->clk);
+               complete(&tspi->xfer_completion);
+               spin_unlock_irqrestore(&tspi->lock, flags);
+               return IRQ_HANDLED;
+       }
+
+       if (tspi->cur_direction & DATA_DIR_RX)
+               tegra_spi_copy_spi_rxbuf_to_client_rxbuf(tspi, t);
+
+       if (tspi->cur_direction & DATA_DIR_TX)
+               tspi->cur_pos = tspi->cur_tx_pos;
+       else
+               tspi->cur_pos = tspi->cur_rx_pos;
+
+       if (tspi->cur_pos == t->len) {
+               complete(&tspi->xfer_completion);
+               goto exit;
+       }
+
+       /* Continue transfer in current message */
+       total_fifo_words = tegra_spi_calculate_curr_xfer_param(tspi->cur_spi,
+                                                       tspi, t);
+       if (total_fifo_words > SPI_FIFO_DEPTH)
+               err = tegra_spi_start_dma_based_transfer(tspi, t);
+       else
+               err = tegra_spi_start_cpu_based_transfer(tspi, t);
+
+exit:
+       spin_unlock_irqrestore(&tspi->lock, flags);
+       return IRQ_HANDLED;
+}
+
+static irqreturn_t tegra_spi_isr_thread(int irq, void *context_data)
+{
+       struct tegra_spi_data *tspi = context_data;
+
+       if (!tspi->is_curr_dma_xfer)
+               return handle_cpu_based_xfer(tspi);
+       return handle_dma_based_xfer(tspi);
+}
+
+static irqreturn_t tegra_spi_isr(int irq, void *context_data)
+{
+       struct tegra_spi_data *tspi = context_data;
+
+       tspi->status_reg = tegra_spi_readl(tspi, SPI_FIFO_STATUS);
+       if (tspi->cur_direction & DATA_DIR_TX)
+               tspi->tx_status = tspi->status_reg &
+                                       (SPI_TX_FIFO_UNF | SPI_TX_FIFO_OVF);
+
+       if (tspi->cur_direction & DATA_DIR_RX)
+               tspi->rx_status = tspi->status_reg &
+                                       (SPI_RX_FIFO_OVF | SPI_RX_FIFO_UNF);
+       tegra_spi_clear_status(tspi);
+
+       return IRQ_WAKE_THREAD;
+}
+
+static void tegra_spi_parse_dt(struct platform_device *pdev,
+       struct tegra_spi_data *tspi)
+{
+       struct device_node *np = pdev->dev.of_node;
+       u32 of_dma[2];
+
+       if (of_property_read_u32_array(np, "nvidia,dma-request-selector",
+                               of_dma, 2) >= 0)
+               tspi->dma_req_sel = of_dma[1];
+
+       if (of_property_read_u32(np, "spi-max-frequency",
+                               &tspi->spi_max_frequency))
+               tspi->spi_max_frequency = 25000000; /* 25MHz */
+}
+
+static struct of_device_id tegra_spi_of_match[] = {
+       { .compatible = "nvidia,tegra114-spi", },
+       {}
+};
+MODULE_DEVICE_TABLE(of, tegra_spi_of_match);
+
+static int tegra_spi_probe(struct platform_device *pdev)
+{
+       struct spi_master       *master;
+       struct tegra_spi_data   *tspi;
+       struct resource         *r;
+       int ret, spi_irq;
+
+       master = spi_alloc_master(&pdev->dev, sizeof(*tspi));
+       if (!master) {
+               dev_err(&pdev->dev, "master allocation failed\n");
+               return -ENOMEM;
+       }
+       dev_set_drvdata(&pdev->dev, master);
+       tspi = spi_master_get_devdata(master);
+
+       /* Parse DT */
+       tegra_spi_parse_dt(pdev, tspi);
+
+       /* the spi->mode bits understood by this driver: */
+       master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
+       master->setup = tegra_spi_setup;
+       master->transfer_one_message = tegra_spi_transfer_one_message;
+       master->num_chipselect = MAX_CHIP_SELECT;
+       master->bus_num = -1;
+
+       tspi->master = master;
+       tspi->dev = &pdev->dev;
+       spin_lock_init(&tspi->lock);
+
+       r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+       if (!r) {
+               dev_err(&pdev->dev, "No IO memory resource\n");
+               ret = -ENODEV;
+               goto exit_free_master;
+       }
+       tspi->phys = r->start;
+       tspi->base = devm_ioremap_resource(&pdev->dev, r);
+       if (IS_ERR(tspi->base)) {
+               ret = PTR_ERR(tspi->base);
+               dev_err(&pdev->dev, "ioremap failed: err = %d\n", ret);
+               goto exit_free_master;
+       }
+
+       spi_irq = platform_get_irq(pdev, 0);
+       tspi->irq = spi_irq;
+       ret = request_threaded_irq(tspi->irq, tegra_spi_isr,
+                       tegra_spi_isr_thread, IRQF_ONESHOT,
+                       dev_name(&pdev->dev), tspi);
+       if (ret < 0) {
+               dev_err(&pdev->dev, "Failed to register ISR for IRQ %d\n",
+                                       tspi->irq);
+               goto exit_free_master;
+       }
+
+       tspi->clk = devm_clk_get(&pdev->dev, "spi");
+       if (IS_ERR(tspi->clk)) {
+               dev_err(&pdev->dev, "can not get clock\n");
+               ret = PTR_ERR(tspi->clk);
+               goto exit_free_irq;
+       }
+
+       tspi->max_buf_size = SPI_FIFO_DEPTH << 2;
+       tspi->dma_buf_size = DEFAULT_SPI_DMA_BUF_LEN;
+
+       if (tspi->dma_req_sel) {
+               ret = tegra_spi_init_dma_param(tspi, true);
+               if (ret < 0) {
+                       dev_err(&pdev->dev, "RxDma Init failed, err %d\n", ret);
+                       goto exit_free_irq;
+               }
+
+               ret = tegra_spi_init_dma_param(tspi, false);
+               if (ret < 0) {
+                       dev_err(&pdev->dev, "TxDma Init failed, err %d\n", ret);
+                       goto exit_rx_dma_free;
+               }
+               tspi->max_buf_size = tspi->dma_buf_size;
+               init_completion(&tspi->tx_dma_complete);
+               init_completion(&tspi->rx_dma_complete);
+       }
+
+       init_completion(&tspi->xfer_completion);
+
+       pm_runtime_enable(&pdev->dev);
+       if (!pm_runtime_enabled(&pdev->dev)) {
+               ret = tegra_spi_runtime_resume(&pdev->dev);
+               if (ret)
+                       goto exit_pm_disable;
+       }
+
+       ret = pm_runtime_get_sync(&pdev->dev);
+       if (ret < 0) {
+               dev_err(&pdev->dev, "pm runtime get failed, e = %d\n", ret);
+               goto exit_pm_disable;
+       }
+       tspi->def_command1_reg  = SPI_M_S;
+       tegra_spi_writel(tspi, tspi->def_command1_reg, SPI_COMMAND1);
+       pm_runtime_put(&pdev->dev);
+
+       master->dev.of_node = pdev->dev.of_node;
+       ret = spi_register_master(master);
+       if (ret < 0) {
+               dev_err(&pdev->dev, "can not register to master err %d\n", ret);
+               goto exit_pm_disable;
+       }
+       return ret;
+
+exit_pm_disable:
+       pm_runtime_disable(&pdev->dev);
+       if (!pm_runtime_status_suspended(&pdev->dev))
+               tegra_spi_runtime_suspend(&pdev->dev);
+       tegra_spi_deinit_dma_param(tspi, false);
+exit_rx_dma_free:
+       tegra_spi_deinit_dma_param(tspi, true);
+exit_free_irq:
+       free_irq(spi_irq, tspi);
+exit_free_master:
+       spi_master_put(master);
+       return ret;
+}
+
+static int tegra_spi_remove(struct platform_device *pdev)
+{
+       struct spi_master *master = dev_get_drvdata(&pdev->dev);
+       struct tegra_spi_data   *tspi = spi_master_get_devdata(master);
+
+       free_irq(tspi->irq, tspi);
+       spi_unregister_master(master);
+
+       if (tspi->tx_dma_chan)
+               tegra_spi_deinit_dma_param(tspi, false);
+
+       if (tspi->rx_dma_chan)
+               tegra_spi_deinit_dma_param(tspi, true);
+
+       pm_runtime_disable(&pdev->dev);
+       if (!pm_runtime_status_suspended(&pdev->dev))
+               tegra_spi_runtime_suspend(&pdev->dev);
+
+       return 0;
+}
+
+#ifdef CONFIG_PM_SLEEP
+static int tegra_spi_suspend(struct device *dev)
+{
+       struct spi_master *master = dev_get_drvdata(dev);
+
+       return spi_master_suspend(master);
+}
+
+static int tegra_spi_resume(struct device *dev)
+{
+       struct spi_master *master = dev_get_drvdata(dev);
+       struct tegra_spi_data *tspi = spi_master_get_devdata(master);
+       int ret;
+
+       ret = pm_runtime_get_sync(dev);
+       if (ret < 0) {
+               dev_err(dev, "pm runtime failed, e = %d\n", ret);
+               return ret;
+       }
+       tegra_spi_writel(tspi, tspi->command1_reg, SPI_COMMAND1);
+       pm_runtime_put(dev);
+
+       return spi_master_resume(master);
+}
+#endif
+
+static int tegra_spi_runtime_suspend(struct device *dev)
+{
+       struct spi_master *master = dev_get_drvdata(dev);
+       struct tegra_spi_data *tspi = spi_master_get_devdata(master);
+
+       /* Flush all write which are in PPSB queue by reading back */
+       tegra_spi_readl(tspi, SPI_COMMAND1);
+
+       clk_disable_unprepare(tspi->clk);
+       return 0;
+}
+
+static int tegra_spi_runtime_resume(struct device *dev)
+{
+       struct spi_master *master = dev_get_drvdata(dev);
+       struct tegra_spi_data *tspi = spi_master_get_devdata(master);
+       int ret;
+
+       ret = clk_prepare_enable(tspi->clk);
+       if (ret < 0) {
+               dev_err(tspi->dev, "clk_prepare failed: %d\n", ret);
+               return ret;
+       }
+       return 0;
+}
+
+static const struct dev_pm_ops tegra_spi_pm_ops = {
+       SET_RUNTIME_PM_OPS(tegra_spi_runtime_suspend,
+               tegra_spi_runtime_resume, NULL)
+       SET_SYSTEM_SLEEP_PM_OPS(tegra_spi_suspend, tegra_spi_resume)
+};
+static struct platform_driver tegra_spi_driver = {
+       .driver = {
+               .name           = "spi-tegra114",
+               .owner          = THIS_MODULE,
+               .pm             = &tegra_spi_pm_ops,
+               .of_match_table = tegra_spi_of_match,
+       },
+       .probe =        tegra_spi_probe,
+       .remove =       tegra_spi_remove,
+};
+module_platform_driver(tegra_spi_driver);
+
+MODULE_ALIAS("platform:spi-tegra114");
+MODULE_DESCRIPTION("NVIDIA Tegra114 SPI Controller Driver");
+MODULE_AUTHOR("Laxman Dewangan <ldewangan@nvidia.com>");
+MODULE_LICENSE("GPL v2");