[PATCH v4 0/9] Add Tegra Quad SPI driver

[PATCH v4 0/9] Add Tegra Quad SPI driver

[Sowjanya Komatineni]

This series adds Tegra210, Tegra186, and Tegra194 Quad SPI driver and
enables Quad SPI on Jetson Nano and Jetson Xavier NX.

QSPI controller is available on Tegra210, Tegra186 and Tegra194.

Tegra186 and Tegra194 has additional feature of combined sequence mode
where command, address and data can all be transferred in a single transfer.
Combined sequence mode is useful only when using DMA mode transfer.

This series does not have combined sequence mode feature as Tegra186/Tegra194
GPCDMA driver is not upstreamed yet.

This series includes
- dt-binding document
- QSPI driver for Tegra210/Tegra186/Tegra194
- Enables QSPI on Jetson Nano and Jetson Xavier NX.

Delta between patch versions:
[v4]:	Updated dummy cycles implementation based on v3 feedback
	- Added dummy_data bit field int spi_transfer to indicate corresponding
	  transfer is dummy bytes transfer.
	- Updated Tegra QSPI transfer_one_message to identify dummy transfer and
	  to use HW supported dummy bytes transfer when dummy cycles are with in
	  Tegra QSPI supported max HW dummy cycles otherwise fallsback to transfer
	  dummy bytes from software.
	- Updated dt-bindings based on v3 feedback.

[v3]:	v2 has some mixed patches sent out accidentally.
	v3 sends proper patches with fixes mentioned in v2.

[v2]:	below v1 feedback
	- Added SPI_MASTER_USES_HW_DUMMY_CYCLES flag for controllers supporting
	  hardware dummy cycles and skips dummy bytes transfer from software for
	  these controllers.
	- Updated dt-binding doc with tx/rx tap delay properties.
	- Added qspi_out clock to dt-binding doc which will be used later with
	  ddr mode support.
	- All other v1 feedback on some cleanup.


Sowjanya Komatineni (9):
  dt-bindings: clock: tegra: Add clock ID TEGRA210_CLK_QSPI_PM
  dt-bindings: spi: Add Tegra Quad SPI device tree binding
  MAINTAINERS: Add Tegra Quad SPI driver section
  spi: tegra210-quad: Add support for Tegra210 QSPI controller
  spi: spi-mem: Mark dummy transfers by setting dummy_data bit
  spi: tegra210-quad: Add support for hardware dummy cycles transfer
  arm64: tegra: Enable QSPI on Jetson Nano
  arm64: tegra: Add QSPI nodes on Tegra194
  arm64: tegra: Enable QSPI on Jetson Xavier NX

 .../bindings/spi/nvidia,tegra210-quad.yaml         |  117 ++
 MAINTAINERS                                        |    8 +
 .../dts/nvidia/tegra194-p3509-0000+p3668-0000.dts  |   12 +
 arch/arm64/boot/dts/nvidia/tegra194.dtsi           |   28 +
 arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts |   12 +
 arch/arm64/boot/dts/nvidia/tegra210.dtsi           |    5 +-
 drivers/spi/Kconfig                                |    9 +
 drivers/spi/Makefile                               |    1 +
 drivers/spi/spi-mem.c                              |    1 +
 drivers/spi/spi-tegra210-quad.c                    | 1421 ++++++++++++++++++++
 include/dt-bindings/clock/tegra210-car.h           |    2 +-
 include/linux/spi/spi.h                            |    2 +
 12 files changed, 1615 insertions(+), 3 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml
 create mode 100644 drivers/spi/spi-tegra210-quad.c

-- 
2.7.4

[PATCH v4 1/9] dt-bindings: clock: tegra: Add clock ID TEGRA210_CLK_QSPI_PM

[Sowjanya Komatineni]

Tegra210 QSPI clock output has divider DIV2_SEL which will be enabled
when using DDR interface mode.

This patch adds clock ID for this to dt-binding.

Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 include/dt-bindings/clock/tegra210-car.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/dt-bindings/clock/tegra210-car.h b/include/dt-bindings/clock/tegra210-car.h
index ab8b8a7..9cfcc3b 100644
--- a/include/dt-bindings/clock/tegra210-car.h
+++ b/include/dt-bindings/clock/tegra210-car.h
@@ -307,7 +307,7 @@
 #define TEGRA210_CLK_AUDIO4 275
 #define TEGRA210_CLK_SPDIF 276
 /* 277 */
-/* 278 */
+#define TEGRA210_CLK_QSPI_PM 278
 /* 279 */
 /* 280 */
 #define TEGRA210_CLK_SOR0_LVDS 281 /* deprecated */
-- 
2.7.4

[PATCH v4 2/9] dt-bindings: spi: Add Tegra Quad SPI device tree binding

[Sowjanya Komatineni]

This patch adds YAML based device tree binding document for Tegra
Quad SPI driver.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 .../bindings/spi/nvidia,tegra210-quad.yaml         | 117 +++++++++++++++++++++
 1 file changed, 117 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml

diff --git a/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml b/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml
new file mode 100644
index 0000000..35a8045
--- /dev/null
+++ b/Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml
@@ -0,0 +1,117 @@
+# SPDX-License-Identifier: (GPL-2.0 OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/spi/nvidia,tegra210-quad.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Tegra Quad SPI Controller
+
+maintainers:
+  - Thierry Reding <thierry.reding@gmail.com>
+  - Jonathan Hunter <jonathanh@nvidia.com>
+
+allOf:
+  - $ref: "spi-controller.yaml#"
+
+properties:
+  compatible:
+    enum:
+      - nvidia,tegra210-qspi
+      - nvidia,tegra186-qspi
+      - nvidia,tegra194-qspi
+
+  reg:
+    maxItems: 1
+
+  interrupts:
+    maxItems: 1
+
+  clock-names:
+    items:
+      - const: qspi
+      - const: qspi_out
+
+  clocks:
+    maxItems: 2
+
+  resets:
+    maxItems: 1
+
+  dmas:
+    maxItems: 2
+
+  dma-names:
+    items:
+      - const: rx
+      - const: tx
+
+patternProperties:
+  "@[0-9a-f]+":
+    type: object
+
+    properties:
+      spi-rx-bus-width:
+        enum: [1, 2, 4]
+
+      spi-tx-bus-width:
+        enum: [1, 2, 4]
+
+      nvidia,tx-clk-tap-delay:
+        description:
+          Delays the clock going out to device with this tap value.
+          Tap value varies based on platform design trace lengths from Tegra
+          QSPI to corresponding slave device.
+        $ref: /schemas/types.yaml#/definitions/uint32
+        minimum: 0
+        maximum: 31
+
+      nvidia,rx-clk-tap-delay:
+        description:
+          Delays the clock coming in from the device with this tap value.
+          Tap value varies based on platform design trace lengths from Tegra
+          QSPI to corresponding slave device.
+        $ref: /schemas/types.yaml#/definitions/uint32
+        minimum: 0
+        maximum: 255
+
+    required:
+      - reg
+
+required:
+  - compatible
+  - reg
+  - interrupts
+  - clock-names
+  - clocks
+  - resets
+
+unevaluatedProperties: false
+
+examples:
+  - |
+    #include <dt-bindings/clock/tegra210-car.h>
+    #include <dt-bindings/reset/tegra210-car.h>
+    #include <dt-bindings/interrupt-controller/arm-gic.h>
+    spi@70410000 {
+            compatible = "nvidia,tegra210-qspi";
+            reg = <0x70410000 0x1000>;
+            interrupts = <GIC_SPI 10 IRQ_TYPE_LEVEL_HIGH>;
+            #address-cells = <1>;
+            #size-cells = <0>;
+            clocks = <&tegra_car TEGRA210_CLK_QSPI>,
+                     <&tegra_car TEGRA210_CLK_QSPI_PM>;
+            clock-names = "qspi", "qspi_out";
+            resets = <&tegra_car 211>;
+            dmas = <&apbdma 5>, <&apbdma 5>;
+            dma-names = "rx", "tx";
+
+            flash@0 {
+                    compatible = "spi-nor";
+                    reg = <0>;
+                    spi-max-frequency = <104000000>;
+                    spi-tx-bus-width = <2>;
+                    spi-rx-bus-width = <2>;
+                    nvidia,tx-clk-tap-delay = <0>;
+                    nvidia,rx-clk-tap-delay = <0>;
+            };
+    };
-- 
2.7.4

Re: [PATCH v4 2/9] dt-bindings: spi: Add Tegra Quad SPI device tree binding

[Rob Herring]

On Thu, 17 Dec 2020 12:28:40 -0800, Sowjanya Komatineni wrote:
> This patch adds YAML based device tree binding document for Tegra
> Quad SPI driver.
> 
> Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
> ---
>  .../bindings/spi/nvidia,tegra210-quad.yaml         | 117 +++++++++++++++++++++
>  1 file changed, 117 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/spi/nvidia,tegra210-quad.yaml
> 

Reviewed-by: Rob Herring <robh@kernel.org>

[PATCH v4 3/9] MAINTAINERS: Add Tegra Quad SPI driver section

[Sowjanya Komatineni]

Add maintainers and mailing list entries to Tegra Quad SPI driver
section.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 MAINTAINERS | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index 5b20bab..19db61f 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -17447,6 +17447,14 @@ M:	Laxman Dewangan <ldewangan@nvidia.com>
 S:	Supported
 F:	drivers/spi/spi-tegra*
 
+TEGRA QUAD SPI DRIVER
+M:	Thierry Reding <thierry.reding@gmail.com>
+M:	Jonathan Hunter <jonathanh@nvidia.com>
+M:	Sowjanya Komatineni <skomatineni@nvidia.com>
+L:	linux-tegra@vger.kernel.org
+S:	Maintained
+F:	drivers/spi/spi-tegra210-quad.c
+
 TEGRA VIDEO DRIVER
 M:	Thierry Reding <thierry.reding@gmail.com>
 M:	Jonathan Hunter <jonathanh@nvidia.com>
-- 
2.7.4

[PATCH v4 4/9] spi: tegra210-quad: Add support for Tegra210 QSPI controller

[Sowjanya Komatineni]

Tegra SoC has a Quad SPI controller starting from Tegra210.

This patch adds support for Tegra210 QSPI controller.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 drivers/spi/Kconfig             |    9 +
 drivers/spi/Makefile            |    1 +
 drivers/spi/spi-tegra210-quad.c | 1382 +++++++++++++++++++++++++++++++++++++++
 3 files changed, 1392 insertions(+)
 create mode 100644 drivers/spi/spi-tegra210-quad.c

diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
index aadaea0..f56f20e 100644
--- a/drivers/spi/Kconfig
+++ b/drivers/spi/Kconfig
@@ -843,6 +843,15 @@ config SPI_MXS
 	help
 	  SPI driver for Freescale MXS devices.
 
+config SPI_TEGRA210_QUAD
+	tristate "NVIDIA Tegra QSPI Controller"
+	depends on ARCH_TEGRA || COMPILE_TEST
+	depends on RESET_CONTROLLER
+	help
+	  QSPI driver for NVIDIA Tegra QSPI Controller interface. This
+	  controller is different from the SPI controller and is available
+	  on Tegra SoCs starting from Tegra210.
+
 config SPI_TEGRA114
 	tristate "NVIDIA Tegra114 SPI Controller"
 	depends on (ARCH_TEGRA && TEGRA20_APB_DMA) || COMPILE_TEST
diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
index 6fea582..c822c5e 100644
--- a/drivers/spi/Makefile
+++ b/drivers/spi/Makefile
@@ -115,6 +115,7 @@ obj-$(CONFIG_SPI_ST_SSC4)		+= spi-st-ssc4.o
 obj-$(CONFIG_SPI_SUN4I)			+= spi-sun4i.o
 obj-$(CONFIG_SPI_SUN6I)			+= spi-sun6i.o
 obj-$(CONFIG_SPI_SYNQUACER)		+= spi-synquacer.o
+obj-$(CONFIG_SPI_TEGRA210_QUAD)		+= spi-tegra210-quad.o
 obj-$(CONFIG_SPI_TEGRA114)		+= spi-tegra114.o
 obj-$(CONFIG_SPI_TEGRA20_SFLASH)	+= spi-tegra20-sflash.o
 obj-$(CONFIG_SPI_TEGRA20_SLINK)		+= spi-tegra20-slink.o
diff --git a/drivers/spi/spi-tegra210-quad.c b/drivers/spi/spi-tegra210-quad.c
new file mode 100644
index 0000000..e7bee8d
--- /dev/null
+++ b/drivers/spi/spi-tegra210-quad.c
@@ -0,0 +1,1382 @@
+// SPDX-License-Identifier: GPL-2.0-only
+//
+// Copyright (C) 2020 NVIDIA CORPORATION.
+
+#include <linux/clk.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/interrupt.h>
+#include <linux/io.h>
+#include <linux/iopoll.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/reset.h>
+#include <linux/spi/spi.h>
+
+#define QSPI_COMMAND1				0x000
+#define QSPI_BIT_LENGTH(x)			(((x) & 0x1f) << 0)
+#define QSPI_PACKED				BIT(5)
+#define QSPI_INTERFACE_WIDTH_MASK		(0x03 << 7)
+#define QSPI_INTERFACE_WIDTH(x)			(((x) & 0x03) << 7)
+#define QSPI_INTERFACE_WIDTH_SINGLE		QSPI_INTERFACE_WIDTH(0)
+#define QSPI_INTERFACE_WIDTH_DUAL		QSPI_INTERFACE_WIDTH(1)
+#define QSPI_INTERFACE_WIDTH_QUAD		QSPI_INTERFACE_WIDTH(2)
+#define QSPI_SDR_DDR_SEL			BIT(9)
+#define QSPI_TX_EN				BIT(11)
+#define QSPI_RX_EN				BIT(12)
+#define QSPI_CS_SW_VAL				BIT(20)
+#define QSPI_CS_SW_HW				BIT(21)
+#define QSPI_CONTROL_MODE_0			(0 << 28)
+#define QSPI_CONTROL_MODE_3			(3 << 28)
+#define QSPI_CONTROL_MODE_MASK			(3 << 28)
+#define QSPI_M_S				BIT(30)
+#define QSPI_PIO				BIT(31)
+
+#define QSPI_COMMAND2				0x004
+#define QSPI_TX_TAP_DELAY(x)			(((x) & 0x3f) << 10)
+#define QSPI_RX_TAP_DELAY(x)			(((x) & 0xff) << 0)
+
+#define QSPI_CS_TIMING1				0x008
+#define QSPI_SETUP_HOLD(setup, hold)		(((setup) << 4) | (hold))
+
+#define QSPI_CS_TIMING2				0x00c
+#define CYCLES_BETWEEN_PACKETS_0(x)		(((x) & 0x1f) << 0)
+#define CS_ACTIVE_BETWEEN_PACKETS_0		BIT(5)
+
+#define QSPI_TRANS_STATUS			0x010
+#define QSPI_BLK_CNT(val)			(((val) >> 0) & 0xffff)
+#define QSPI_RDY				BIT(30)
+
+#define QSPI_FIFO_STATUS			0x014
+#define QSPI_RX_FIFO_EMPTY			BIT(0)
+#define QSPI_RX_FIFO_FULL			BIT(1)
+#define QSPI_TX_FIFO_EMPTY			BIT(2)
+#define QSPI_TX_FIFO_FULL			BIT(3)
+#define QSPI_RX_FIFO_UNF			BIT(4)
+#define QSPI_RX_FIFO_OVF			BIT(5)
+#define QSPI_TX_FIFO_UNF			BIT(6)
+#define QSPI_TX_FIFO_OVF			BIT(7)
+#define QSPI_ERR				BIT(8)
+#define QSPI_TX_FIFO_FLUSH			BIT(14)
+#define QSPI_RX_FIFO_FLUSH			BIT(15)
+#define QSPI_TX_FIFO_EMPTY_COUNT(val)		(((val) >> 16) & 0x7f)
+#define QSPI_RX_FIFO_FULL_COUNT(val)		(((val) >> 23) & 0x7f)
+
+#define QSPI_FIFO_ERROR				(QSPI_RX_FIFO_UNF | \
+						 QSPI_RX_FIFO_OVF | \
+						 QSPI_TX_FIFO_UNF | \
+						 QSPI_TX_FIFO_OVF)
+#define QSPI_FIFO_EMPTY				(QSPI_RX_FIFO_EMPTY | \
+						 QSPI_TX_FIFO_EMPTY)
+
+#define QSPI_TX_DATA				0x018
+#define QSPI_RX_DATA				0x01c
+
+#define QSPI_DMA_CTL				0x020
+#define QSPI_TX_TRIG(n)				(((n) & 0x3) << 15)
+#define QSPI_TX_TRIG_1				QSPI_TX_TRIG(0)
+#define QSPI_TX_TRIG_4				QSPI_TX_TRIG(1)
+#define QSPI_TX_TRIG_8				QSPI_TX_TRIG(2)
+#define QSPI_TX_TRIG_16				QSPI_TX_TRIG(3)
+
+#define QSPI_RX_TRIG(n)				(((n) & 0x3) << 19)
+#define QSPI_RX_TRIG_1				QSPI_RX_TRIG(0)
+#define QSPI_RX_TRIG_4				QSPI_RX_TRIG(1)
+#define QSPI_RX_TRIG_8				QSPI_RX_TRIG(2)
+#define QSPI_RX_TRIG_16				QSPI_RX_TRIG(3)
+
+#define QSPI_DMA_EN				BIT(31)
+
+#define QSPI_DMA_BLK				0x024
+#define QSPI_DMA_BLK_SET(x)			(((x) & 0xffff) << 0)
+
+#define QSPI_TX_FIFO				0x108
+#define QSPI_RX_FIFO				0x188
+
+#define QSPI_FIFO_DEPTH				64
+
+#define QSPI_INTR_MASK				0x18c
+#define QSPI_INTR_RX_FIFO_UNF_MASK		BIT(25)
+#define QSPI_INTR_RX_FIFO_OVF_MASK		BIT(26)
+#define QSPI_INTR_TX_FIFO_UNF_MASK		BIT(27)
+#define QSPI_INTR_TX_FIFO_OVF_MASK		BIT(28)
+#define QSPI_INTR_RDY_MASK			BIT(29)
+#define QSPI_INTR_RX_TX_FIFO_ERR		(QSPI_INTR_RX_FIFO_UNF_MASK | \
+						 QSPI_INTR_RX_FIFO_OVF_MASK | \
+						 QSPI_INTR_TX_FIFO_UNF_MASK | \
+						 QSPI_INTR_TX_FIFO_OVF_MASK)
+
+#define QSPI_MISC_REG                           0x194
+#define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
+
+#define DATA_DIR_TX				BIT(0)
+#define DATA_DIR_RX				BIT(1)
+
+#define QSPI_DMA_TIMEOUT			(msecs_to_jiffies(1000))
+#define DEFAULT_QSPI_DMA_BUF_LEN		(64 * 1024)
+
+struct tegra_qspi_client_data {
+	int tx_clk_tap_delay;
+	int rx_clk_tap_delay;
+};
+
+struct tegra_qspi {
+	struct device				*dev;
+	struct spi_master			*master;
+	/* lock to protect data accessed by irq */
+	spinlock_t				lock;
+
+	struct clk				*clk;
+	struct reset_control			*rst;
+	void __iomem				*base;
+	phys_addr_t				phys;
+	unsigned int				irq;
+
+	u32					cur_speed;
+	unsigned int				cur_pos;
+	unsigned int				words_per_32bit;
+	unsigned int				bytes_per_word;
+	unsigned int				curr_dma_words;
+	unsigned int				cur_direction;
+
+	unsigned int				cur_rx_pos;
+	unsigned int				cur_tx_pos;
+
+	unsigned int				dma_buf_size;
+	unsigned int				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;
+	bool					use_dma;
+
+	u32					command1_reg;
+	u32					dma_control_reg;
+	u32					def_command1_reg;
+	u32					def_command2_reg;
+	u32					spi_cs_timing1;
+	u32					spi_cs_timing2;
+
+	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 inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
+{
+	return readl(tqspi->base + offset);
+}
+
+static inline void tegra_qspi_writel(struct tegra_qspi *tqspi, u32 value, unsigned long offset)
+{
+	writel(value, tqspi->base + offset);
+
+	/* read back register to make sure that register writes completed */
+	if (offset != QSPI_TX_FIFO)
+		readl(tqspi->base + QSPI_COMMAND1);
+}
+
+static void tegra_qspi_mask_clear_irq(struct tegra_qspi *tqspi)
+{
+	u32 value;
+
+	/* write 1 to clear status register */
+	value = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
+	tegra_qspi_writel(tqspi, value, QSPI_TRANS_STATUS);
+
+	value = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
+	if (!(value & QSPI_INTR_RDY_MASK)) {
+		value |= (QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
+		tegra_qspi_writel(tqspi, value, QSPI_INTR_MASK);
+	}
+
+	/* clear fifo status error if any */
+	value = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
+	if (value & QSPI_ERR)
+		tegra_qspi_writel(tqspi, QSPI_ERR | QSPI_FIFO_ERROR, QSPI_FIFO_STATUS);
+}
+
+static unsigned int
+tegra_qspi_calculate_curr_xfer_param(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	unsigned int max_word, max_len, total_fifo_words;
+	unsigned int remain_len = t->len - tqspi->cur_pos;
+	unsigned int bits_per_word = t->bits_per_word;
+
+	tqspi->bytes_per_word = DIV_ROUND_UP(bits_per_word, 8);
+
+	/*
+	 * Tegra QSPI controller supports packed or unpacked mode transfers.
+	 * Packed mode is used for data transfers using 8, 16, or 32 bits per
+	 * word with a minimum transfer of 1 word and for all other transfers
+	 * unpacked mode will be used.
+	 */
+
+	if ((bits_per_word == 8 || bits_per_word == 16 ||
+	     bits_per_word == 32) && t->len > 3) {
+		tqspi->is_packed = true;
+		tqspi->words_per_32bit = 32 / bits_per_word;
+	} else {
+		tqspi->is_packed = false;
+		tqspi->words_per_32bit = 1;
+	}
+
+	if (tqspi->is_packed) {
+		max_len = min(remain_len, tqspi->max_buf_size);
+		tqspi->curr_dma_words = max_len / tqspi->bytes_per_word;
+		total_fifo_words = (max_len + 3) / 4;
+	} else {
+		max_word = (remain_len - 1) / tqspi->bytes_per_word + 1;
+		max_word = min(max_word, tqspi->max_buf_size / 4);
+		tqspi->curr_dma_words = max_word;
+		total_fifo_words = max_word;
+	}
+
+	return total_fifo_words;
+}
+
+static unsigned int
+tegra_qspi_fill_tx_fifo_from_client_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	unsigned int written_words, fifo_words_left, count;
+	unsigned int len, tx_empty_count, max_n_32bit, i;
+	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
+	u32 fifo_status;
+
+	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
+	tx_empty_count = QSPI_TX_FIFO_EMPTY_COUNT(fifo_status);
+
+	if (tqspi->is_packed) {
+		fifo_words_left = tx_empty_count * tqspi->words_per_32bit;
+		written_words = min(fifo_words_left, tqspi->curr_dma_words);
+		len = written_words * tqspi->bytes_per_word;
+		max_n_32bit = DIV_ROUND_UP(len, 4);
+		for (count = 0; count < max_n_32bit; count++) {
+			u32 x = 0;
+
+			for (i = 0; (i < 4) && len; i++, len--)
+				x |= (u32)(*tx_buf++) << (i * 8);
+			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
+		}
+
+		tqspi->cur_tx_pos += written_words * tqspi->bytes_per_word;
+	} else {
+		unsigned int write_bytes;
+		u8 bytes_per_word = tqspi->bytes_per_word;
+
+		max_n_32bit = min(tqspi->curr_dma_words, tx_empty_count);
+		written_words = max_n_32bit;
+		len = written_words * tqspi->bytes_per_word;
+		if (len > t->len - tqspi->cur_pos)
+			len = t->len - tqspi->cur_pos;
+		write_bytes = len;
+		for (count = 0; count < max_n_32bit; count++) {
+			u32 x = 0;
+
+			for (i = 0; len && (i < bytes_per_word); i++, len--)
+				x |= (u32)(*tx_buf++) << (i * 8);
+			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
+		}
+
+		tqspi->cur_tx_pos += write_bytes;
+	}
+
+	return written_words;
+}
+
+static unsigned int
+tegra_qspi_read_rx_fifo_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
+	unsigned int len, rx_full_count, count, i;
+	unsigned int read_words = 0;
+	u32 fifo_status, x;
+
+	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
+	rx_full_count = QSPI_RX_FIFO_FULL_COUNT(fifo_status);
+	if (tqspi->is_packed) {
+		len = tqspi->curr_dma_words * tqspi->bytes_per_word;
+		for (count = 0; count < rx_full_count; count++) {
+			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO);
+
+			for (i = 0; len && (i < 4); i++, len--)
+				*rx_buf++ = (x >> i * 8) & 0xff;
+		}
+
+		read_words += tqspi->curr_dma_words;
+		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
+	} else {
+		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
+		u8 bytes_per_word = tqspi->bytes_per_word;
+		unsigned int read_bytes;
+
+		len = rx_full_count * bytes_per_word;
+		if (len > t->len - tqspi->cur_pos)
+			len = t->len - tqspi->cur_pos;
+		read_bytes = len;
+		for (count = 0; count < rx_full_count; count++) {
+			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO) & rx_mask;
+
+			for (i = 0; len && (i < bytes_per_word); i++, len--)
+				*rx_buf++ = (x >> (i * 8)) & 0xff;
+		}
+
+		read_words += rx_full_count;
+		tqspi->cur_rx_pos += read_bytes;
+	}
+
+	return read_words;
+}
+
+static void
+tegra_qspi_copy_client_txbuf_to_qspi_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	dma_sync_single_for_cpu(tqspi->dev, tqspi->tx_dma_phys,
+				tqspi->dma_buf_size, DMA_TO_DEVICE);
+
+	/*
+	 * In packed mode, each word in FIFO may contain multiple packets
+	 * based on bits per word. So all bytes in each FIFO word are valid.
+	 *
+	 * In unpacked mode, each word in FIFO contains single packet and
+	 * based on bits per word any remaining bits in FIFO word will be
+	 * ignored by the hardware and are invalid bits.
+	 */
+	if (tqspi->is_packed) {
+		tqspi->cur_tx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
+	} else {
+		u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
+		unsigned int i, count, consume, write_bytes;
+
+		/*
+		 * Fill tx_dma_buf to contain single packet in each word based
+		 * on bits per word from SPI core tx_buf.
+		 */
+		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
+		if (consume > t->len - tqspi->cur_pos)
+			consume = t->len - tqspi->cur_pos;
+		write_bytes = consume;
+		for (count = 0; count < tqspi->curr_dma_words; count++) {
+			u32 x = 0;
+
+			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
+				x |= (u32)(*tx_buf++) << (i * 8);
+			tqspi->tx_dma_buf[count] = x;
+		}
+
+		tqspi->cur_tx_pos += write_bytes;
+	}
+
+	dma_sync_single_for_device(tqspi->dev, tqspi->tx_dma_phys,
+				   tqspi->dma_buf_size, DMA_TO_DEVICE);
+}
+
+static void
+tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	dma_sync_single_for_cpu(tqspi->dev, tqspi->rx_dma_phys,
+				tqspi->dma_buf_size, DMA_FROM_DEVICE);
+
+	if (tqspi->is_packed) {
+		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
+	} else {
+		unsigned char *rx_buf = t->rx_buf + tqspi->cur_rx_pos;
+		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
+		unsigned int i, count, consume, read_bytes;
+
+		/*
+		 * Each FIFO word contains single data packet.
+		 * Skip invalid bits in each FIFO word based on bits per word
+		 * and align bytes while filling in SPI core rx_buf.
+		 */
+		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
+		if (consume > t->len - tqspi->cur_pos)
+			consume = t->len - tqspi->cur_pos;
+		read_bytes = consume;
+		for (count = 0; count < tqspi->curr_dma_words; count++) {
+			u32 x = tqspi->rx_dma_buf[count] & rx_mask;
+
+			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
+				*rx_buf++ = (x >> (i * 8)) & 0xff;
+		}
+
+		tqspi->cur_rx_pos += read_bytes;
+	}
+
+	dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
+				   tqspi->dma_buf_size, DMA_FROM_DEVICE);
+}
+
+static void tegra_qspi_dma_complete(void *args)
+{
+	struct completion *dma_complete = args;
+
+	complete(dma_complete);
+}
+
+static int tegra_qspi_start_tx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
+{
+	dma_addr_t tx_dma_phys;
+
+	reinit_completion(&tqspi->tx_dma_complete);
+
+	if (tqspi->is_packed)
+		tx_dma_phys = t->tx_dma;
+	else
+		tx_dma_phys = tqspi->tx_dma_phys;
+
+	tqspi->tx_dma_desc = dmaengine_prep_slave_single(tqspi->tx_dma_chan, tx_dma_phys,
+							 len, DMA_MEM_TO_DEV,
+							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
+
+	if (!tqspi->tx_dma_desc) {
+		dev_err(tqspi->dev, "Unable to get TX descriptor\n");
+		return -EIO;
+	}
+
+	tqspi->tx_dma_desc->callback = tegra_qspi_dma_complete;
+	tqspi->tx_dma_desc->callback_param = &tqspi->tx_dma_complete;
+	dmaengine_submit(tqspi->tx_dma_desc);
+	dma_async_issue_pending(tqspi->tx_dma_chan);
+
+	return 0;
+}
+
+static int tegra_qspi_start_rx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
+{
+	dma_addr_t rx_dma_phys;
+
+	reinit_completion(&tqspi->rx_dma_complete);
+
+	if (tqspi->is_packed)
+		rx_dma_phys = t->rx_dma;
+	else
+		rx_dma_phys = tqspi->rx_dma_phys;
+
+	tqspi->rx_dma_desc = dmaengine_prep_slave_single(tqspi->rx_dma_chan, rx_dma_phys,
+							 len, DMA_DEV_TO_MEM,
+							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);
+
+	if (!tqspi->rx_dma_desc) {
+		dev_err(tqspi->dev, "Unable to get RX descriptor\n");
+		return -EIO;
+	}
+
+	tqspi->rx_dma_desc->callback = tegra_qspi_dma_complete;
+	tqspi->rx_dma_desc->callback_param = &tqspi->rx_dma_complete;
+	dmaengine_submit(tqspi->rx_dma_desc);
+	dma_async_issue_pending(tqspi->rx_dma_chan);
+
+	return 0;
+}
+
+static int tegra_qspi_flush_fifos(struct tegra_qspi *tqspi, bool atomic)
+{
+	void __iomem *addr = tqspi->base + QSPI_FIFO_STATUS;
+	u32 val;
+
+	val = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
+	if ((val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY)
+		return 0;
+
+	val |= QSPI_RX_FIFO_FLUSH | QSPI_TX_FIFO_FLUSH;
+	tegra_qspi_writel(tqspi, val, QSPI_FIFO_STATUS);
+
+	if (!atomic)
+		return readl_relaxed_poll_timeout(addr, val,
+						  (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
+						  1000, 1000000);
+
+	return readl_relaxed_poll_timeout_atomic(addr, val,
+						 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
+						 1000, 1000000);
+}
+
+static void tegra_qspi_unmask_irq(struct tegra_qspi *tqspi)
+{
+	u32 intr_mask;
+
+	intr_mask = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
+	intr_mask &= ~(QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
+	tegra_qspi_writel(tqspi, intr_mask, QSPI_INTR_MASK);
+}
+
+static int tegra_qspi_dma_map_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
+	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
+	unsigned int len;
+
+	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
+
+	if (t->tx_buf) {
+		t->tx_dma = dma_map_single(tqspi->dev, (void *)tx_buf, len, DMA_TO_DEVICE);
+		if (dma_mapping_error(tqspi->dev, t->tx_dma))
+			return -ENOMEM;
+	}
+
+	if (t->rx_buf) {
+		t->rx_dma = dma_map_single(tqspi->dev, (void *)rx_buf, len, DMA_FROM_DEVICE);
+		if (dma_mapping_error(tqspi->dev, t->rx_dma)) {
+			dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
+			return -ENOMEM;
+		}
+	}
+
+	return 0;
+}
+
+static void tegra_qspi_dma_unmap_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	unsigned int len;
+
+	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
+
+	dma_unmap_single(tqspi->dev, t->tx_dma, len, DMA_TO_DEVICE);
+	dma_unmap_single(tqspi->dev, t->rx_dma, len, DMA_FROM_DEVICE);
+}
+
+static int tegra_qspi_start_dma_based_transfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
+{
+	struct dma_slave_config dma_sconfig = { 0 };
+	unsigned int len;
+	u8 dma_burst;
+	int ret = 0;
+	u32 val;
+
+	if (tqspi->is_packed) {
+		ret = tegra_qspi_dma_map_xfer(tqspi, t);
+		if (ret < 0)
+			return ret;
+	}
+
+	val = QSPI_DMA_BLK_SET(tqspi->curr_dma_words - 1);
+	tegra_qspi_writel(tqspi, val, QSPI_DMA_BLK);
+
+	tegra_qspi_unmask_irq(tqspi);
+
+	if (tqspi->is_packed)
+		len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
+	else
+		len = tqspi->curr_dma_words * 4;
+
+	/* set attention level based on length of transfer */
+	val = 0;
+	if (len & 0xf) {
+		val |= QSPI_TX_TRIG_1 | QSPI_RX_TRIG_1;
+		dma_burst = 1;
+	} else if (((len) >> 4) & 0x1) {
+		val |= QSPI_TX_TRIG_4 | QSPI_RX_TRIG_4;
+		dma_burst = 4;
+	} else {
+		val |= QSPI_TX_TRIG_8 | QSPI_RX_TRIG_8;
+		dma_burst = 8;
+	}
+
+	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
+	tqspi->dma_control_reg = val;
+
+	dma_sconfig.device_fc = true;
+	if (tqspi->cur_direction & DATA_DIR_TX) {
+		dma_sconfig.dst_addr = tqspi->phys + QSPI_TX_FIFO;
+		dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+		dma_sconfig.dst_maxburst = dma_burst;
+		ret = dmaengine_slave_config(tqspi->tx_dma_chan, &dma_sconfig);
+		if (ret < 0) {
+			dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
+			return ret;
+		}
+
+		tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
+		ret = tegra_qspi_start_tx_dma(tqspi, t, len);
+		if (ret < 0) {
+			dev_err(tqspi->dev, "failed to starting TX DMA: %d\n", ret);
+			return ret;
+		}
+	}
+
+	if (tqspi->cur_direction & DATA_DIR_RX) {
+		dma_sconfig.src_addr = tqspi->phys + QSPI_RX_FIFO;
+		dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+		dma_sconfig.src_maxburst = dma_burst;
+		ret = dmaengine_slave_config(tqspi->rx_dma_chan, &dma_sconfig);
+		if (ret < 0) {
+			dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
+			return ret;
+		}
+
+		dma_sync_single_for_device(tqspi->dev, tqspi->rx_dma_phys,
+					   tqspi->dma_buf_size,
+					   DMA_FROM_DEVICE);
+
+		ret = tegra_qspi_start_rx_dma(tqspi, t, len);
+		if (ret < 0) {
+			dev_err(tqspi->dev, "failed to start RX DMA: %d\n", ret);
+			if (tqspi->cur_direction & DATA_DIR_TX)
+				dmaengine_terminate_all(tqspi->tx_dma_chan);
+			return ret;
+		}
+	}
+
+	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
+
+	tqspi->is_curr_dma_xfer = true;
+	tqspi->dma_control_reg = val;
+	val |= QSPI_DMA_EN;
+	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
+
+	return ret;
+}
+
+static int tegra_qspi_start_cpu_based_transfer(struct tegra_qspi *qspi, struct spi_transfer *t)
+{
+	u32 val;
+	unsigned int cur_words;
+
+	if (qspi->cur_direction & DATA_DIR_TX)
+		cur_words = tegra_qspi_fill_tx_fifo_from_client_txbuf(qspi, t);
+	else
+		cur_words = qspi->curr_dma_words;
+
+	val = QSPI_DMA_BLK_SET(cur_words - 1);
+	tegra_qspi_writel(qspi, val, QSPI_DMA_BLK);
+
+	tegra_qspi_unmask_irq(qspi);
+
+	qspi->is_curr_dma_xfer = false;
+	val = qspi->command1_reg;
+	val |= QSPI_PIO;
+	tegra_qspi_writel(qspi, val, QSPI_COMMAND1);
+
+	return 0;
+}
+
+static void tegra_qspi_deinit_dma(struct tegra_qspi *tqspi)
+{
+	if (tqspi->tx_dma_buf) {
+		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
+				  tqspi->tx_dma_buf, tqspi->tx_dma_phys);
+		tqspi->tx_dma_buf = NULL;
+	}
+
+	if (tqspi->tx_dma_chan) {
+		dma_release_channel(tqspi->tx_dma_chan);
+		tqspi->tx_dma_chan = NULL;
+	}
+
+	if (tqspi->rx_dma_buf) {
+		dma_free_coherent(tqspi->dev, tqspi->dma_buf_size,
+				  tqspi->rx_dma_buf, tqspi->rx_dma_phys);
+		tqspi->rx_dma_buf = NULL;
+	}
+
+	if (tqspi->rx_dma_chan) {
+		dma_release_channel(tqspi->rx_dma_chan);
+		tqspi->rx_dma_chan = NULL;
+	}
+}
+
+static int tegra_qspi_init_dma(struct tegra_qspi *tqspi)
+{
+	struct dma_chan *dma_chan;
+	dma_addr_t dma_phys;
+	u32 *dma_buf;
+	int err;
+
+	dma_chan = dma_request_chan(tqspi->dev, "rx");
+	if (IS_ERR(dma_chan)) {
+		err = PTR_ERR(dma_chan);
+		goto err_out;
+	}
+
+	tqspi->rx_dma_chan = dma_chan;
+
+	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
+	if (!dma_buf) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+
+	tqspi->rx_dma_buf = dma_buf;
+	tqspi->rx_dma_phys = dma_phys;
+
+	dma_chan = dma_request_chan(tqspi->dev, "tx");
+	if (IS_ERR(dma_chan)) {
+		err = PTR_ERR(dma_chan);
+		goto err_out;
+	}
+
+	tqspi->tx_dma_chan = dma_chan;
+
+	dma_buf = dma_alloc_coherent(tqspi->dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
+	if (!dma_buf) {
+		err = -ENOMEM;
+		goto err_out;
+	}
+
+	tqspi->tx_dma_buf = dma_buf;
+	tqspi->tx_dma_phys = dma_phys;
+	tqspi->use_dma = true;
+
+	return 0;
+
+err_out:
+	tegra_qspi_deinit_dma(tqspi);
+
+	if (err != -EPROBE_DEFER) {
+		dev_err(tqspi->dev, "cannot use DMA: %d\n", err);
+		dev_err(tqspi->dev, "falling back to PIO\n");
+		return 0;
+	}
+
+	return err;
+}
+
+static u32 tegra_qspi_setup_transfer_one(struct spi_device *spi, struct spi_transfer *t,
+					 bool is_first_of_msg)
+{
+	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
+	struct tegra_qspi_client_data *cdata = spi->controller_data;
+	u32 command1, command2, speed = t->speed_hz;
+	u8 bits_per_word = t->bits_per_word;
+	u32 tx_tap = 0, rx_tap = 0;
+	int req_mode;
+
+	if (speed != tqspi->cur_speed) {
+		clk_set_rate(tqspi->clk, speed);
+		tqspi->cur_speed = speed;
+	}
+
+	tqspi->cur_pos = 0;
+	tqspi->cur_rx_pos = 0;
+	tqspi->cur_tx_pos = 0;
+	tqspi->curr_xfer = t;
+
+	if (is_first_of_msg) {
+		tegra_qspi_mask_clear_irq(tqspi);
+
+		command1 = tqspi->def_command1_reg;
+		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
+
+		command1 &= ~QSPI_CONTROL_MODE_MASK;
+		req_mode = spi->mode & 0x3;
+		if (req_mode == SPI_MODE_3)
+			command1 |= QSPI_CONTROL_MODE_3;
+		else
+			command1 |= QSPI_CONTROL_MODE_0;
+
+		if (spi->mode & SPI_CS_HIGH)
+			command1 |= QSPI_CS_SW_VAL;
+		else
+			command1 &= ~QSPI_CS_SW_VAL;
+		tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
+
+		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 = QSPI_TX_TAP_DELAY(tx_tap) | QSPI_RX_TAP_DELAY(rx_tap);
+		if (command2 != tqspi->def_command2_reg)
+			tegra_qspi_writel(tqspi, command2, QSPI_COMMAND2);
+
+	} else {
+		command1 = tqspi->command1_reg;
+		command1 &= ~QSPI_BIT_LENGTH(~0);
+		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
+	}
+
+	command1 &= ~QSPI_SDR_DDR_SEL;
+
+	return command1;
+}
+
+static int tegra_qspi_start_transfer_one(struct spi_device *spi,
+					 struct spi_transfer *t, u32 command1)
+{
+	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
+	unsigned int total_fifo_words;
+	u8 bus_width = 0;
+	int ret;
+
+	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
+
+	command1 &= ~QSPI_PACKED;
+	if (tqspi->is_packed)
+		command1 |= QSPI_PACKED;
+	tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);
+
+	tqspi->cur_direction = 0;
+
+	command1 &= ~(QSPI_TX_EN | QSPI_RX_EN);
+	if (t->rx_buf) {
+		command1 |= QSPI_RX_EN;
+		tqspi->cur_direction |= DATA_DIR_RX;
+		bus_width = t->rx_nbits;
+	}
+
+	if (t->tx_buf) {
+		command1 |= QSPI_TX_EN;
+		tqspi->cur_direction |= DATA_DIR_TX;
+		bus_width = t->tx_nbits;
+	}
+
+	command1 &= ~QSPI_INTERFACE_WIDTH_MASK;
+
+	if (bus_width == SPI_NBITS_QUAD)
+		command1 |= QSPI_INTERFACE_WIDTH_QUAD;
+	else if (bus_width == SPI_NBITS_DUAL)
+		command1 |= QSPI_INTERFACE_WIDTH_DUAL;
+	else
+		command1 |= QSPI_INTERFACE_WIDTH_SINGLE;
+
+	tqspi->command1_reg = command1;
+
+	ret = tegra_qspi_flush_fifos(tqspi, false);
+	if (ret < 0)
+		return ret;
+
+	if (tqspi->use_dma && total_fifo_words > QSPI_FIFO_DEPTH)
+		ret = tegra_qspi_start_dma_based_transfer(tqspi, t);
+	else
+		ret = tegra_qspi_start_cpu_based_transfer(tqspi, t);
+
+	return ret;
+}
+
+static struct tegra_qspi_client_data *tegra_qspi_parse_cdata_dt(struct spi_device *spi)
+{
+	struct tegra_qspi_client_data *cdata;
+	struct device_node *slave_np = spi->dev.of_node;
+
+	cdata = kzalloc(sizeof(*cdata), GFP_KERNEL);
+	if (!cdata)
+		return NULL;
+
+	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;
+}
+
+static void tegra_qspi_cleanup(struct spi_device *spi)
+{
+	struct tegra_qspi_client_data *cdata = spi->controller_data;
+
+	spi->controller_data = NULL;
+	kfree(cdata);
+}
+
+static int tegra_qspi_setup(struct spi_device *spi)
+{
+	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
+	struct tegra_qspi_client_data *cdata = spi->controller_data;
+	unsigned long flags;
+	u32 val;
+	int ret;
+
+	ret = pm_runtime_resume_and_get(tqspi->dev);
+	if (ret < 0) {
+		dev_err(tqspi->dev, "failed to get runtime PM: %d\n", ret);
+		return ret;
+	}
+
+	if (!cdata) {
+		cdata = tegra_qspi_parse_cdata_dt(spi);
+		spi->controller_data = cdata;
+	}
+
+	spin_lock_irqsave(&tqspi->lock, flags);
+
+	/* keep default cs state to inactive */
+	val = tqspi->def_command1_reg;
+	if (spi->mode & SPI_CS_HIGH)
+		val &= ~QSPI_CS_SW_VAL;
+	else
+		val |= QSPI_CS_SW_VAL;
+
+	tqspi->def_command1_reg = val;
+	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
+
+	spin_unlock_irqrestore(&tqspi->lock, flags);
+
+	pm_runtime_put(tqspi->dev);
+
+	return 0;
+}
+
+static void tegra_qspi_dump_regs(struct tegra_qspi *tqspi)
+{
+	dev_dbg(tqspi->dev, "============ QSPI REGISTER DUMP ============\n");
+	dev_dbg(tqspi->dev, "Command1:    0x%08x | Command2:    0x%08x\n",
+		tegra_qspi_readl(tqspi, QSPI_COMMAND1),
+		tegra_qspi_readl(tqspi, QSPI_COMMAND2));
+	dev_dbg(tqspi->dev, "DMA_CTL:     0x%08x | DMA_BLK:     0x%08x\n",
+		tegra_qspi_readl(tqspi, QSPI_DMA_CTL),
+		tegra_qspi_readl(tqspi, QSPI_DMA_BLK));
+	dev_dbg(tqspi->dev, "INTR_MASK:  0x%08x | MISC: 0x%08x\n",
+		tegra_qspi_readl(tqspi, QSPI_INTR_MASK),
+		tegra_qspi_readl(tqspi, QSPI_MISC_REG));
+	dev_dbg(tqspi->dev, "TRANS_STAT:  0x%08x | FIFO_STATUS: 0x%08x\n",
+		tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS),
+		tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS));
+}
+
+static void tegra_qspi_handle_error(struct tegra_qspi *tqspi)
+{
+	dev_err(tqspi->dev, "error in transfer, fifo status 0x%08x\n", tqspi->status_reg);
+	tegra_qspi_dump_regs(tqspi);
+	tegra_qspi_flush_fifos(tqspi, true);
+	reset_control_assert(tqspi->rst);
+	udelay(2);
+	reset_control_deassert(tqspi->rst);
+}
+
+static void tegra_qspi_transfer_end(struct spi_device *spi)
+{
+	struct tegra_qspi *tqspi = spi_master_get_devdata(spi->master);
+	int cs_val = (spi->mode & SPI_CS_HIGH) ? 0 : 1;
+
+	if (cs_val)
+		tqspi->command1_reg |= QSPI_CS_SW_VAL;
+	else
+		tqspi->command1_reg &= ~QSPI_CS_SW_VAL;
+	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
+	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
+}
+
+static int tegra_qspi_transfer_one_message(struct spi_master *master, struct spi_message *msg)
+{
+	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
+	struct spi_device *spi = msg->spi;
+	struct spi_transfer *xfer;
+	bool is_first_msg = true;
+	int ret;
+
+	msg->status = 0;
+	msg->actual_length = 0;
+	tqspi->tx_status = 0;
+	tqspi->rx_status = 0;
+
+	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
+		u32 cmd1;
+
+		reinit_completion(&tqspi->xfer_completion);
+
+		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
+
+		ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
+		if (ret < 0) {
+			dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
+			goto complete_xfer;
+		}
+
+		is_first_msg = false;
+		ret = wait_for_completion_timeout(&tqspi->xfer_completion,
+						  QSPI_DMA_TIMEOUT);
+		if (WARN_ON(ret == 0)) {
+			dev_err(tqspi->dev, "transfer timeout: %d\n", ret);
+			if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_TX))
+				dmaengine_terminate_all(tqspi->tx_dma_chan);
+			if (tqspi->is_curr_dma_xfer && (tqspi->cur_direction & DATA_DIR_RX))
+				dmaengine_terminate_all(tqspi->rx_dma_chan);
+			tegra_qspi_handle_error(tqspi);
+			ret = -EIO;
+			goto complete_xfer;
+		}
+
+		if (tqspi->tx_status ||  tqspi->rx_status) {
+			tegra_qspi_handle_error(tqspi);
+			ret = -EIO;
+			goto complete_xfer;
+		}
+
+		msg->actual_length += xfer->len;
+
+complete_xfer:
+		if (ret < 0) {
+			tegra_qspi_transfer_end(spi);
+			spi_transfer_delay_exec(xfer);
+			goto exit;
+		}
+
+		if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
+			/* de-activate CS after last transfer only when cs_change is not set */
+			if (!xfer->cs_change) {
+				tegra_qspi_transfer_end(spi);
+				spi_transfer_delay_exec(xfer);
+			}
+		} else if (xfer->cs_change) {
+			 /* de-activated CS between the transfers only when cs_change is set */
+			tegra_qspi_transfer_end(spi);
+			spi_transfer_delay_exec(xfer);
+		}
+	}
+
+	ret = 0;
+exit:
+	msg->status = ret;
+	spi_finalize_current_message(master);
+	return ret;
+}
+
+static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
+{
+	struct spi_transfer *t = tqspi->curr_xfer;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tqspi->lock, flags);
+
+	if (tqspi->tx_status ||  tqspi->rx_status) {
+		tegra_qspi_handle_error(tqspi);
+		complete(&tqspi->xfer_completion);
+		goto exit;
+	}
+
+	if (tqspi->cur_direction & DATA_DIR_RX)
+		tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);
+
+	if (tqspi->cur_direction & DATA_DIR_TX)
+		tqspi->cur_pos = tqspi->cur_tx_pos;
+	else
+		tqspi->cur_pos = tqspi->cur_rx_pos;
+
+	if (tqspi->cur_pos == t->len) {
+		complete(&tqspi->xfer_completion);
+		goto exit;
+	}
+
+	tegra_qspi_calculate_curr_xfer_param(tqspi, t);
+	tegra_qspi_start_cpu_based_transfer(tqspi, t);
+exit:
+	spin_unlock_irqrestore(&tqspi->lock, flags);
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
+{
+	struct spi_transfer *t = tqspi->curr_xfer;
+	unsigned int total_fifo_words;
+	unsigned long flags;
+	long wait_status;
+	int err = 0;
+
+	if (tqspi->cur_direction & DATA_DIR_TX) {
+		if (tqspi->tx_status) {
+			dmaengine_terminate_all(tqspi->tx_dma_chan);
+			err += 1;
+		} else {
+			wait_status = wait_for_completion_interruptible_timeout(
+				&tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
+			if (wait_status <= 0) {
+				dmaengine_terminate_all(tqspi->tx_dma_chan);
+				dev_err(tqspi->dev, "failed TX DMA transfer\n");
+				err += 1;
+			}
+		}
+	}
+
+	if (tqspi->cur_direction & DATA_DIR_RX) {
+		if (tqspi->rx_status) {
+			dmaengine_terminate_all(tqspi->rx_dma_chan);
+			err += 2;
+		} else {
+			wait_status = wait_for_completion_interruptible_timeout(
+				&tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
+			if (wait_status <= 0) {
+				dmaengine_terminate_all(tqspi->rx_dma_chan);
+				dev_err(tqspi->dev, "failed RX DMA transfer\n");
+				err += 2;
+			}
+		}
+	}
+
+	spin_lock_irqsave(&tqspi->lock, flags);
+
+	if (err) {
+		tegra_qspi_dma_unmap_xfer(tqspi, t);
+		tegra_qspi_handle_error(tqspi);
+		complete(&tqspi->xfer_completion);
+		goto exit;
+	}
+
+	if (tqspi->cur_direction & DATA_DIR_RX)
+		tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);
+
+	if (tqspi->cur_direction & DATA_DIR_TX)
+		tqspi->cur_pos = tqspi->cur_tx_pos;
+	else
+		tqspi->cur_pos = tqspi->cur_rx_pos;
+
+	if (tqspi->cur_pos == t->len) {
+		tegra_qspi_dma_unmap_xfer(tqspi, t);
+		complete(&tqspi->xfer_completion);
+		goto exit;
+	}
+
+	tegra_qspi_dma_unmap_xfer(tqspi, t);
+
+	/* continue transfer in current message */
+	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
+	if (total_fifo_words > QSPI_FIFO_DEPTH)
+		err = tegra_qspi_start_dma_based_transfer(tqspi, t);
+	else
+		err = tegra_qspi_start_cpu_based_transfer(tqspi, t);
+
+exit:
+	spin_unlock_irqrestore(&tqspi->lock, flags);
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t tegra_qspi_isr_thread(int irq, void *context_data)
+{
+	struct tegra_qspi *tqspi = context_data;
+
+	tqspi->status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
+
+	if (tqspi->cur_direction & DATA_DIR_TX)
+		tqspi->tx_status = tqspi->status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
+
+	if (tqspi->cur_direction & DATA_DIR_RX)
+		tqspi->rx_status = tqspi->status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
+
+	tegra_qspi_mask_clear_irq(tqspi);
+
+	if (!tqspi->is_curr_dma_xfer)
+		return handle_cpu_based_xfer(tqspi);
+
+	return handle_dma_based_xfer(tqspi);
+}
+
+static const struct of_device_id tegra_qspi_of_match[] = {
+	{ .compatible = "nvidia,tegra210-qspi", },
+	{ .compatible = "nvidia,tegra186-qspi", },
+	{ .compatible = "nvidia,tegra194-qspi", },
+	{}
+};
+
+MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);
+
+static int tegra_qspi_probe(struct platform_device *pdev)
+{
+	struct spi_master	*master;
+	struct tegra_qspi	*tqspi;
+	struct resource		*r;
+	int ret, qspi_irq;
+	int bus_num;
+
+	master = devm_spi_alloc_master(&pdev->dev, sizeof(*tqspi));
+	if (!master)
+		return -ENOMEM;
+
+	platform_set_drvdata(pdev, master);
+	tqspi = spi_master_get_devdata(master);
+
+	master->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
+			    SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
+	master->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
+	master->setup = tegra_qspi_setup;
+	master->cleanup = tegra_qspi_cleanup;
+	master->transfer_one_message = tegra_qspi_transfer_one_message;
+	master->num_chipselect = 1;
+	master->auto_runtime_pm = true;
+
+	bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
+	if (bus_num >= 0)
+		master->bus_num = bus_num;
+
+	tqspi->master = master;
+	tqspi->dev = &pdev->dev;
+	spin_lock_init(&tqspi->lock);
+
+	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	tqspi->base = devm_ioremap_resource(&pdev->dev, r);
+	if (IS_ERR(tqspi->base))
+		return PTR_ERR(tqspi->base);
+
+	tqspi->phys = r->start;
+	qspi_irq = platform_get_irq(pdev, 0);
+	tqspi->irq = qspi_irq;
+
+	tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
+	if (IS_ERR(tqspi->clk)) {
+		ret = PTR_ERR(tqspi->clk);
+		dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
+		return ret;
+	}
+
+	tqspi->rst = devm_reset_control_get_exclusive(&pdev->dev, NULL);
+	if (IS_ERR(tqspi->rst)) {
+		ret = PTR_ERR(tqspi->rst);
+		dev_err(&pdev->dev, "failed to get reset control: %d\n", ret);
+		return ret;
+	}
+
+	tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
+	tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;
+
+	ret = tegra_qspi_init_dma(tqspi);
+	if (ret < 0)
+		return ret;
+
+	if (tqspi->use_dma)
+		tqspi->max_buf_size = tqspi->dma_buf_size;
+
+	init_completion(&tqspi->tx_dma_complete);
+	init_completion(&tqspi->rx_dma_complete);
+	init_completion(&tqspi->xfer_completion);
+
+	pm_runtime_enable(&pdev->dev);
+	ret = pm_runtime_resume_and_get(&pdev->dev);
+	if (ret < 0) {
+		dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
+		goto exit_pm_disable;
+	}
+
+	reset_control_assert(tqspi->rst);
+	udelay(2);
+	reset_control_deassert(tqspi->rst);
+
+	tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW |  QSPI_CS_SW_VAL;
+	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
+	tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
+	tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
+	tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);
+
+	pm_runtime_put(&pdev->dev);
+
+	ret = request_threaded_irq(tqspi->irq, NULL,
+				   tegra_qspi_isr_thread, IRQF_ONESHOT,
+				   dev_name(&pdev->dev), tqspi);
+	if (ret < 0) {
+		dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
+		goto exit_pm_disable;
+	}
+
+	master->dev.of_node = pdev->dev.of_node;
+	ret = spi_register_master(master);
+	if (ret < 0) {
+		dev_err(&pdev->dev, "failed to register master: %d\n", ret);
+		goto exit_free_irq;
+	}
+
+	return 0;
+
+exit_free_irq:
+	free_irq(qspi_irq, tqspi);
+exit_pm_disable:
+	pm_runtime_disable(&pdev->dev);
+	tegra_qspi_deinit_dma(tqspi);
+	return ret;
+}
+
+static int tegra_qspi_remove(struct platform_device *pdev)
+{
+	struct spi_master *master = platform_get_drvdata(pdev);
+	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
+
+	spi_unregister_master(master);
+	free_irq(tqspi->irq, tqspi);
+	pm_runtime_disable(&pdev->dev);
+	tegra_qspi_deinit_dma(tqspi);
+
+	return 0;
+}
+
+static int __maybe_unused tegra_qspi_suspend(struct device *dev)
+{
+	struct spi_master *master = dev_get_drvdata(dev);
+
+	return spi_master_suspend(master);
+}
+
+static int __maybe_unused tegra_qspi_resume(struct device *dev)
+{
+	struct spi_master *master = dev_get_drvdata(dev);
+	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
+	int ret;
+
+	ret = pm_runtime_resume_and_get(dev);
+	if (ret < 0) {
+		dev_err(dev, "failed to get runtime PM: %d\n", ret);
+		return ret;
+	}
+
+	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
+	tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
+	pm_runtime_put(dev);
+
+	return spi_master_resume(master);
+}
+
+static int __maybe_unused tegra_qspi_runtime_suspend(struct device *dev)
+{
+	struct spi_master *master = dev_get_drvdata(dev);
+	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
+
+	/* flush all write which are in PPSB queue by reading back */
+	tegra_qspi_readl(tqspi, QSPI_COMMAND1);
+
+	clk_disable_unprepare(tqspi->clk);
+
+	return 0;
+}
+
+static int __maybe_unused tegra_qspi_runtime_resume(struct device *dev)
+{
+	struct spi_master *master = dev_get_drvdata(dev);
+	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
+	int ret;
+
+	ret = clk_prepare_enable(tqspi->clk);
+	if (ret < 0)
+		dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);
+
+	return ret;
+}
+
+static const struct dev_pm_ops tegra_qspi_pm_ops = {
+	SET_RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
+	SET_SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
+};
+
+static struct platform_driver tegra_qspi_driver = {
+	.driver = {
+		.name		= "tegra-qspi",
+		.pm		= &tegra_qspi_pm_ops,
+		.of_match_table	= tegra_qspi_of_match,
+	},
+	.probe =	tegra_qspi_probe,
+	.remove =	tegra_qspi_remove,
+};
+module_platform_driver(tegra_qspi_driver);
+
+MODULE_ALIAS("platform:qspi-tegra");
+MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
+MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
+MODULE_LICENSE("GPL v2");
-- 
2.7.4

[PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Sowjanya Komatineni]

This patch marks dummy transfer by setting dummy_data bit to 1.

Controllers supporting dummy transfer by hardware use this bit field
to skip software transfer of dummy bytes and use hardware dummy bytes
transfer.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 drivers/spi/spi-mem.c   | 1 +
 include/linux/spi/spi.h | 2 ++
 2 files changed, 3 insertions(+)

diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c
index f3a3f19..c64371c 100644
--- a/drivers/spi/spi-mem.c
+++ b/drivers/spi/spi-mem.c
@@ -354,6 +354,7 @@ int spi_mem_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
 		xfers[xferpos].tx_buf = tmpbuf + op->addr.nbytes + 1;
 		xfers[xferpos].len = op->dummy.nbytes;
 		xfers[xferpos].tx_nbits = op->dummy.buswidth;
+		xfers[xferpos].dummy_data = 1;
 		spi_message_add_tail(&xfers[xferpos], &msg);
 		xferpos++;
 		totalxferlen += op->dummy.nbytes;
diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h
index aa09fdc..708f2f5 100644
--- a/include/linux/spi/spi.h
+++ b/include/linux/spi/spi.h
@@ -827,6 +827,7 @@ extern void spi_res_release(struct spi_controller *ctlr,
  *      transfer. If 0 the default (from @spi_device) is used.
  * @bits_per_word: select a bits_per_word other than the device default
  *      for this transfer. If 0 the default (from @spi_device) is used.
+ * @dummy_data: indicates transfer is dummy bytes transfer.
  * @cs_change: affects chipselect after this transfer completes
  * @cs_change_delay: delay between cs deassert and assert when
  *      @cs_change is set and @spi_transfer is not the last in @spi_message
@@ -939,6 +940,7 @@ struct spi_transfer {
 	struct sg_table tx_sg;
 	struct sg_table rx_sg;
 
+	unsigned	dummy_data:1;
 	unsigned	cs_change:1;
 	unsigned	tx_nbits:3;
 	unsigned	rx_nbits:3;
-- 
2.7.4

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Pratyush Yadav]

Hi Sowjanya,

On 17/12/20 12:28PM, Sowjanya Komatineni wrote:
> This patch marks dummy transfer by setting dummy_data bit to 1.
> 
> Controllers supporting dummy transfer by hardware use this bit field
> to skip software transfer of dummy bytes and use hardware dummy bytes
> transfer.

What is the benefit you get from this change? You add complexity in 
spi-mem and the controller driver, so that must come with some benefits. 
Here I don't see any. The transfer will certainly take the same amount 
of time because the number or period of the dummy cycles has not 
changed. So why is this needed?
 
> Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
> ---
>  drivers/spi/spi-mem.c   | 1 +
>  include/linux/spi/spi.h | 2 ++
>  2 files changed, 3 insertions(+)
> 
> diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c
> index f3a3f19..c64371c 100644
> --- a/drivers/spi/spi-mem.c
> +++ b/drivers/spi/spi-mem.c
> @@ -354,6 +354,7 @@ int spi_mem_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
>  		xfers[xferpos].tx_buf = tmpbuf + op->addr.nbytes + 1;
>  		xfers[xferpos].len = op->dummy.nbytes;
>  		xfers[xferpos].tx_nbits = op->dummy.buswidth;
> +		xfers[xferpos].dummy_data = 1;
>  		spi_message_add_tail(&xfers[xferpos], &msg);
>  		xferpos++;
>  		totalxferlen += op->dummy.nbytes;
> diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h
> index aa09fdc..708f2f5 100644
> --- a/include/linux/spi/spi.h
> +++ b/include/linux/spi/spi.h
> @@ -827,6 +827,7 @@ extern void spi_res_release(struct spi_controller *ctlr,
>   *      transfer. If 0 the default (from @spi_device) is used.
>   * @bits_per_word: select a bits_per_word other than the device default
>   *      for this transfer. If 0 the default (from @spi_device) is used.
> + * @dummy_data: indicates transfer is dummy bytes transfer.
>   * @cs_change: affects chipselect after this transfer completes
>   * @cs_change_delay: delay between cs deassert and assert when
>   *      @cs_change is set and @spi_transfer is not the last in @spi_message
> @@ -939,6 +940,7 @@ struct spi_transfer {
>  	struct sg_table tx_sg;
>  	struct sg_table rx_sg;
>  
> +	unsigned	dummy_data:1;
>  	unsigned	cs_change:1;
>  	unsigned	tx_nbits:3;
>  	unsigned	rx_nbits:3;
> -- 
> 2.7.4
> 

-- 
Regards,
Pratyush Yadav
Texas Instruments India

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Boris Brezillon]

On Fri, 18 Dec 2020 14:51:08 +0530
Pratyush Yadav <p.yadav@ti.com> wrote:

> Hi Sowjanya,
> 
> On 17/12/20 12:28PM, Sowjanya Komatineni wrote:
> > This patch marks dummy transfer by setting dummy_data bit to 1.
> > 
> > Controllers supporting dummy transfer by hardware use this bit field
> > to skip software transfer of dummy bytes and use hardware dummy bytes
> > transfer.  
> 
> What is the benefit you get from this change? You add complexity in 
> spi-mem and the controller driver, so that must come with some benefits. 
> Here I don't see any. The transfer will certainly take the same amount 
> of time because the number or period of the dummy cycles has not 
> changed. So why is this needed?

Well, you don't have to queue TX bytes if you use HW-based dummy
cycles, but I agree, I'd expect the overhead to be negligible,
especially since we're talking about emitting a few bytes, not hundreds.
This being said, the complexity added to the core is reasonable IMHO,
so if it really helps reducing the CPU overhead (we might need some
numbers to prove that), I guess it's okay.

>  
> > Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
> > ---
> >  drivers/spi/spi-mem.c   | 1 +
> >  include/linux/spi/spi.h | 2 ++
> >  2 files changed, 3 insertions(+)
> > 
> > diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c
> > index f3a3f19..c64371c 100644
> > --- a/drivers/spi/spi-mem.c
> > +++ b/drivers/spi/spi-mem.c
> > @@ -354,6 +354,7 @@ int spi_mem_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
> >  		xfers[xferpos].tx_buf = tmpbuf + op->addr.nbytes + 1;
> >  		xfers[xferpos].len = op->dummy.nbytes;
> >  		xfers[xferpos].tx_nbits = op->dummy.buswidth;
> > +		xfers[xferpos].dummy_data = 1;
> >  		spi_message_add_tail(&xfers[xferpos], &msg);
> >  		xferpos++;
> >  		totalxferlen += op->dummy.nbytes;
> > diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h
> > index aa09fdc..708f2f5 100644
> > --- a/include/linux/spi/spi.h
> > +++ b/include/linux/spi/spi.h
> > @@ -827,6 +827,7 @@ extern void spi_res_release(struct spi_controller *ctlr,
> >   *      transfer. If 0 the default (from @spi_device) is used.
> >   * @bits_per_word: select a bits_per_word other than the device default
> >   *      for this transfer. If 0 the default (from @spi_device) is used.
> > + * @dummy_data: indicates transfer is dummy bytes transfer.
> >   * @cs_change: affects chipselect after this transfer completes
> >   * @cs_change_delay: delay between cs deassert and assert when
> >   *      @cs_change is set and @spi_transfer is not the last in @spi_message
> > @@ -939,6 +940,7 @@ struct spi_transfer {
> >  	struct sg_table tx_sg;
> >  	struct sg_table rx_sg;
> >  
> > +	unsigned	dummy_data:1;
> >  	unsigned	cs_change:1;
> >  	unsigned	tx_nbits:3;
> >  	unsigned	rx_nbits:3;
> > -- 
> > 2.7.4
> >   
> 

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Sowjanya Komatineni]

On 12/18/20 1:57 AM, Boris Brezillon wrote:
> On Fri, 18 Dec 2020 14:51:08 +0530
> Pratyush Yadav <p.yadav@ti.com> wrote:
>
>> Hi Sowjanya,
>>
>> On 17/12/20 12:28PM, Sowjanya Komatineni wrote:
>>> This patch marks dummy transfer by setting dummy_data bit to 1.
>>>
>>> Controllers supporting dummy transfer by hardware use this bit field
>>> to skip software transfer of dummy bytes and use hardware dummy bytes
>>> transfer.
>> What is the benefit you get from this change? You add complexity in
>> spi-mem and the controller driver, so that must come with some benefits.
>> Here I don't see any. The transfer will certainly take the same amount
>> of time because the number or period of the dummy cycles has not
>> changed. So why is this needed?
> Well, you don't have to queue TX bytes if you use HW-based dummy
> cycles, but I agree, I'd expect the overhead to be negligible,
> especially since we're talking about emitting a few bytes, not hundreds.
> This being said, the complexity added to the core is reasonable IMHO,
> so if it really helps reducing the CPU overhead (we might need some
> numbers to prove that), I guess it's okay.

Hardware dummy cycles feature of Tegra QSPI is to save SW transfer cycle 
of dummy bytes by filling FIFO.

I don't have numbers as we always use hardware dummy cycles with Tegra QSPI.

>>   
>>> Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
>>> ---
>>>   drivers/spi/spi-mem.c   | 1 +
>>>   include/linux/spi/spi.h | 2 ++
>>>   2 files changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c
>>> index f3a3f19..c64371c 100644
>>> --- a/drivers/spi/spi-mem.c
>>> +++ b/drivers/spi/spi-mem.c
>>> @@ -354,6 +354,7 @@ int spi_mem_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
>>>   		xfers[xferpos].tx_buf = tmpbuf + op->addr.nbytes + 1;
>>>   		xfers[xferpos].len = op->dummy.nbytes;
>>>   		xfers[xferpos].tx_nbits = op->dummy.buswidth;
>>> +		xfers[xferpos].dummy_data = 1;
>>>   		spi_message_add_tail(&xfers[xferpos], &msg);
>>>   		xferpos++;
>>>   		totalxferlen += op->dummy.nbytes;
>>> diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h
>>> index aa09fdc..708f2f5 100644
>>> --- a/include/linux/spi/spi.h
>>> +++ b/include/linux/spi/spi.h
>>> @@ -827,6 +827,7 @@ extern void spi_res_release(struct spi_controller *ctlr,
>>>    *      transfer. If 0 the default (from @spi_device) is used.
>>>    * @bits_per_word: select a bits_per_word other than the device default
>>>    *      for this transfer. If 0 the default (from @spi_device) is used.
>>> + * @dummy_data: indicates transfer is dummy bytes transfer.
>>>    * @cs_change: affects chipselect after this transfer completes
>>>    * @cs_change_delay: delay between cs deassert and assert when
>>>    *      @cs_change is set and @spi_transfer is not the last in @spi_message
>>> @@ -939,6 +940,7 @@ struct spi_transfer {
>>>   	struct sg_table tx_sg;
>>>   	struct sg_table rx_sg;
>>>   
>>> +	unsigned	dummy_data:1;
>>>   	unsigned	cs_change:1;
>>>   	unsigned	tx_nbits:3;
>>>   	unsigned	rx_nbits:3;
>>> -- 
>>> 2.7.4
>>>    

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Pratyush Yadav]

On 18/12/20 10:09AM, Sowjanya Komatineni wrote:
> 
> On 12/18/20 1:57 AM, Boris Brezillon wrote:
> > On Fri, 18 Dec 2020 14:51:08 +0530
> > Pratyush Yadav <p.yadav@ti.com> wrote:
> > 
> > > Hi Sowjanya,
> > > 
> > > On 17/12/20 12:28PM, Sowjanya Komatineni wrote:
> > > > This patch marks dummy transfer by setting dummy_data bit to 1.
> > > > 
> > > > Controllers supporting dummy transfer by hardware use this bit field
> > > > to skip software transfer of dummy bytes and use hardware dummy bytes
> > > > transfer.
> > > What is the benefit you get from this change? You add complexity in
> > > spi-mem and the controller driver, so that must come with some benefits.
> > > Here I don't see any. The transfer will certainly take the same amount
> > > of time because the number or period of the dummy cycles has not
> > > changed. So why is this needed?
> > Well, you don't have to queue TX bytes if you use HW-based dummy
> > cycles, but I agree, I'd expect the overhead to be negligible,
> > especially since we're talking about emitting a few bytes, not hundreds.
> > This being said, the complexity added to the core is reasonable IMHO,
> > so if it really helps reducing the CPU overhead (we might need some
> > numbers to prove that), I guess it's okay.
> 
> Hardware dummy cycles feature of Tegra QSPI is to save SW transfer cycle of
> dummy bytes by filling FIFO.
> 
> I don't have numbers as we always use hardware dummy cycles with Tegra QSPI.

Most flashes use somewhere around 8-10 dummy cycles. Some of the faster 
ones working at 166 MHz or above use 20-25. From what I understand, the 
only time hardware transfer of dummy cycles will reduce CPU load 
significantly will be when multiple small transfers are made in quick 
succession. Unless you have such a usecase I don't know how much benefit 
you will get by it.

Anyway, if the SPI maintainers think this is worth it, I won't object.
 
> > > > Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
> > > > ---
> > > >   drivers/spi/spi-mem.c   | 1 +
> > > >   include/linux/spi/spi.h | 2 ++
> > > >   2 files changed, 3 insertions(+)
> > > > 
> > > > diff --git a/drivers/spi/spi-mem.c b/drivers/spi/spi-mem.c
> > > > index f3a3f19..c64371c 100644
> > > > --- a/drivers/spi/spi-mem.c
> > > > +++ b/drivers/spi/spi-mem.c
> > > > @@ -354,6 +354,7 @@ int spi_mem_exec_op(struct spi_mem *mem, const struct spi_mem_op *op)
> > > >   		xfers[xferpos].tx_buf = tmpbuf + op->addr.nbytes + 1;
> > > >   		xfers[xferpos].len = op->dummy.nbytes;
> > > >   		xfers[xferpos].tx_nbits = op->dummy.buswidth;
> > > > +		xfers[xferpos].dummy_data = 1;
> > > >   		spi_message_add_tail(&xfers[xferpos], &msg);
> > > >   		xferpos++;
> > > >   		totalxferlen += op->dummy.nbytes;
> > > > diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h
> > > > index aa09fdc..708f2f5 100644
> > > > --- a/include/linux/spi/spi.h
> > > > +++ b/include/linux/spi/spi.h
> > > > @@ -827,6 +827,7 @@ extern void spi_res_release(struct spi_controller *ctlr,
> > > >    *      transfer. If 0 the default (from @spi_device) is used.
> > > >    * @bits_per_word: select a bits_per_word other than the device default
> > > >    *      for this transfer. If 0 the default (from @spi_device) is used.
> > > > + * @dummy_data: indicates transfer is dummy bytes transfer.
> > > >    * @cs_change: affects chipselect after this transfer completes
> > > >    * @cs_change_delay: delay between cs deassert and assert when
> > > >    *      @cs_change is set and @spi_transfer is not the last in @spi_message
> > > > @@ -939,6 +940,7 @@ struct spi_transfer {
> > > >   	struct sg_table tx_sg;
> > > >   	struct sg_table rx_sg;
> > > > +	unsigned	dummy_data:1;
> > > >   	unsigned	cs_change:1;
> > > >   	unsigned	tx_nbits:3;
> > > >   	unsigned	rx_nbits:3;
> > > > -- 
> > > > 2.7.4

-- 
Regards,
Pratyush Yadav
Texas Instruments India

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 375 bytes --]

On Sat, Dec 19, 2020 at 12:49:38AM +0530, Pratyush Yadav wrote:

> Anyway, if the SPI maintainers think this is worth it, I won't object.

This gets kind of circular, for me it's a question of if there's some
meaningful benefit from using the feature vs the cost to support it and
from the sounds of it we don't have numbers on the benefits from using
it at present.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 848 bytes --]

On Fri, Dec 18, 2020 at 08:41:02PM +0000, Mark Brown wrote:
> On Sat, Dec 19, 2020 at 12:49:38AM +0530, Pratyush Yadav wrote:

> > Anyway, if the SPI maintainers think this is worth it, I won't object.

> This gets kind of circular, for me it's a question of if there's some
> meaningful benefit from using the feature vs the cost to support it and
> from the sounds of it we don't have numbers on the benefits from using
> it at present.

...although I do have to say looking at the implementation that the cost
seems low, it's just a flag set on an existing transfer.  The only issue
is if we'd get more win from coalesing the entire transaction (or entire
transmit) into a single transfer that could be DMAed and/or requires
fewer trips through the stack which does make it seem like an unclear
tradeoff from the point of view of client drivers

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Sowjanya Komatineni]

On 12/18/20 12:44 PM, Mark Brown wrote:
> On Fri, Dec 18, 2020 at 08:41:02PM +0000, Mark Brown wrote:
>> On Sat, Dec 19, 2020 at 12:49:38AM +0530, Pratyush Yadav wrote:
>>> Anyway, if the SPI maintainers think this is worth it, I won't object.
>> This gets kind of circular, for me it's a question of if there's some
>> meaningful benefit from using the feature vs the cost to support it and
>> from the sounds of it we don't have numbers on the benefits from using
>> it at present.
> ...although I do have to say looking at the implementation that the cost
> seems low, it's just a flag set on an existing transfer.  The only issue
> is if we'd get more win from coalesing the entire transaction (or entire
> transmit) into a single transfer that could be DMAed and/or requires
> fewer trips through the stack which does make it seem like an unclear
> tradeoff from the point of view of client drivers

Using HW dummy cycles save extra software cycle of transfer which 
involves transfer setup register writes, writing dummy bytes to TX FIFO, 
interrupt processing.

Implementation wise it just a single bit field added to spi_transfer and 
on Tegra controller driver programming dummy cycles with prior transfer 
and skipping sw dummy transfer which is actually not complex.

 From quick check, I see HW dummy cycles transfer of 128KB shows 18 Mb/s 
while SW transfer of bytes shows 17.3 MB/s on average.

When back-to-back read commands are executed using HW dummy cycles will 
definitely save cycles.

Re: [PATCH v4 5/9] spi: spi-mem: Mark dummy transfers by setting dummy_data bit

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 293 bytes --]

On Fri, Dec 18, 2020 at 02:01:56PM -0800, Sowjanya Komatineni wrote:

> From quick check, I see HW dummy cycles transfer of 128KB shows 18 Mb/s
> while SW transfer of bytes shows 17.3 MB/s on average.

OK, it's not going to revolutionize the world or anything but that's
definitely a speedup.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH v4 6/9] spi: tegra210-quad: Add support for hardware dummy cycles transfer

[Sowjanya Komatineni]

Tegra Quad SPI controller hardware supports sending dummy bytes based
on programmed dummy clock cycles after the actual transfer bytes.

This patch adds this support of hardware dummy bytes transfer and
skips transfer of dummy bytes from the software.

For dummy cycles more than Tegra Quad SPI hardware maximum dummy
cycles limit, driver transfers dummy bytes from the software.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 drivers/spi/spi-tegra210-quad.c | 41 ++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 40 insertions(+), 1 deletion(-)

diff --git a/drivers/spi/spi-tegra210-quad.c b/drivers/spi/spi-tegra210-quad.c
index e7bee8d..695a296 100644
--- a/drivers/spi/spi-tegra210-quad.c
+++ b/drivers/spi/spi-tegra210-quad.c
@@ -117,6 +117,7 @@
 
 #define QSPI_MISC_REG                           0x194
 #define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
+#define QSPI_DUMMY_CYCLES_MAX			0xff
 
 #define DATA_DIR_TX				BIT(0)
 #define DATA_DIR_RX				BIT(1)
@@ -170,6 +171,7 @@ struct tegra_qspi {
 	u32					def_command2_reg;
 	u32					spi_cs_timing1;
 	u32					spi_cs_timing2;
+	u8					dummy_cycles;
 
 	struct completion			xfer_completion;
 	struct spi_transfer			*curr_xfer;
@@ -856,6 +858,8 @@ static int tegra_qspi_start_transfer_one(struct spi_device *spi,
 
 	tqspi->command1_reg = command1;
 
+	tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);
+
 	ret = tegra_qspi_flush_fifos(tqspi, false);
 	if (ret < 0)
 		return ret;
@@ -974,7 +978,8 @@ static int tegra_qspi_transfer_one_message(struct spi_master *master, struct spi
 {
 	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
 	struct spi_device *spi = msg->spi;
-	struct spi_transfer *xfer;
+	struct spi_transfer *xfer, *next_xfer;
+	bool use_hw_dummy_cycles = false;
 	bool is_first_msg = true;
 	int ret;
 
@@ -984,8 +989,42 @@ static int tegra_qspi_transfer_one_message(struct spi_master *master, struct spi
 	tqspi->rx_status = 0;
 
 	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
+		u8 dummy_cycles = 0;
 		u32 cmd1;
 
+		/*
+		 * Skip dummy bytes transfer if they are transferred by the hardware along
+		 * with previous transfer.
+		 */
+		if (xfer->dummy_data && use_hw_dummy_cycles) {
+			msg->actual_length += xfer->len;
+			continue;
+		}
+
+		/*
+		 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
+		 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
+		 * So, check if the next transfer is dummy data transfer and program dummy
+		 * clock cycles along with the current transfer.
+		 */
+		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
+			next_xfer = list_next_entry(xfer, transfer_list);
+			if (next_xfer && next_xfer->dummy_data) {
+				dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
+				use_hw_dummy_cycles = true;
+				/*
+				 * Use software dummy bytes transfer if dummy cycles exceeds
+				 * Tegra QSPI hardware maximum dummy cycles limit.
+				 */
+				if (dummy_cycles > QSPI_DUMMY_CYCLES_MAX) {
+					use_hw_dummy_cycles = false;
+					dummy_cycles = 0;
+				}
+			}
+		}
+
+		tqspi->dummy_cycles = dummy_cycles;
+
 		reinit_completion(&tqspi->xfer_completion);
 
 		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);
-- 
2.7.4

Re: [PATCH v4 6/9] spi: tegra210-quad: Add support for hardware dummy cycles transfer

[Boris Brezillon]

On Thu, 17 Dec 2020 12:28:44 -0800
Sowjanya Komatineni <skomatineni@nvidia.com> wrote:

> Tegra Quad SPI controller hardware supports sending dummy bytes based
> on programmed dummy clock cycles after the actual transfer bytes.
> 
> This patch adds this support of hardware dummy bytes transfer and
> skips transfer of dummy bytes from the software.
> 
> For dummy cycles more than Tegra Quad SPI hardware maximum dummy
> cycles limit, driver transfers dummy bytes from the software.
> 
> Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
> ---
>  drivers/spi/spi-tegra210-quad.c | 41 ++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 40 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/spi/spi-tegra210-quad.c b/drivers/spi/spi-tegra210-quad.c
> index e7bee8d..695a296 100644
> --- a/drivers/spi/spi-tegra210-quad.c
> +++ b/drivers/spi/spi-tegra210-quad.c
> @@ -117,6 +117,7 @@
>  
>  #define QSPI_MISC_REG                           0x194
>  #define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
> +#define QSPI_DUMMY_CYCLES_MAX			0xff
>  
>  #define DATA_DIR_TX				BIT(0)
>  #define DATA_DIR_RX				BIT(1)
> @@ -170,6 +171,7 @@ struct tegra_qspi {
>  	u32					def_command2_reg;
>  	u32					spi_cs_timing1;
>  	u32					spi_cs_timing2;
> +	u8					dummy_cycles;
>  
>  	struct completion			xfer_completion;
>  	struct spi_transfer			*curr_xfer;
> @@ -856,6 +858,8 @@ static int tegra_qspi_start_transfer_one(struct spi_device *spi,
>  
>  	tqspi->command1_reg = command1;
>  
> +	tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);
> +
>  	ret = tegra_qspi_flush_fifos(tqspi, false);
>  	if (ret < 0)
>  		return ret;
> @@ -974,7 +978,8 @@ static int tegra_qspi_transfer_one_message(struct spi_master *master, struct spi
>  {
>  	struct tegra_qspi *tqspi = spi_master_get_devdata(master);
>  	struct spi_device *spi = msg->spi;
> -	struct spi_transfer *xfer;
> +	struct spi_transfer *xfer, *next_xfer;

next_after should be declared where it's actually used.

> +	bool use_hw_dummy_cycles = false;

I don't think you need this variable (see below).

>  	bool is_first_msg = true;
>  	int ret;
>  
> @@ -984,8 +989,42 @@ static int tegra_qspi_transfer_one_message(struct spi_master *master, struct spi
>  	tqspi->rx_status = 0;
>  
>  	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
> +		u8 dummy_cycles = 0;

Should be declared where it's actually used, and you don't want it to be a u8
since you're checking that the result does not exceed 255 which will always
be true with a u8.

>  		u32 cmd1;
>  
> +		/*
> +		 * Skip dummy bytes transfer if they are transferred by the hardware along
> +		 * with previous transfer.
> +		 */
> +		if (xfer->dummy_data && use_hw_dummy_cycles) {
> +			msg->actual_length += xfer->len;
> +			continue;
> +		}
> +
> +		/*
> +		 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
> +		 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
> +		 * So, check if the next transfer is dummy data transfer and program dummy
> +		 * clock cycles along with the current transfer.
> +		 */
> +		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
> +			next_xfer = list_next_entry(xfer, transfer_list);
> +			if (next_xfer && next_xfer->dummy_data) {
> +				dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;
> +				use_hw_dummy_cycles = true;
> +				/*
> +				 * Use software dummy bytes transfer if dummy cycles exceeds
> +				 * Tegra QSPI hardware maximum dummy cycles limit.
> +				 */
> +				if (dummy_cycles > QSPI_DUMMY_CYCLES_MAX) {
> +					use_hw_dummy_cycles = false;
> +					dummy_cycles = 0;
> +				}
> +			}
> +		}
> +
> +		tqspi->dummy_cycles = dummy_cycles;
> +

This can be simplified:

		/*
		 * Skip dummy bytes transfer if they were issued with the
		 * previous transfer.
		 */
		if (tqspi->dummy_cycles) {
			WARN_ON(!xfer->dummy_data);
			tqspi->dummy_cycles = 0;
		}

		/*
		 * Tegra QSPI hardware supports dummy bytes transfer after actual
		 * transfer bytes based on programmed dummy clock cycles in the
		 * QSPI_MISC register. So, check if the next transfer is dummy
		 * data transfer and program dummy clock cycles along with the
		 * current transfer.
		 */
		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
			struct spi_transfer *next_xfer;

			next_xfer = list_next_entry(xfer, transfer_list);
			if (next_xfer->dummy_data) {
				u32 dummy_cycles = next_xfer->len * 8 /
						   next_xfer->tx_nbits;
				if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX)
					tqspi->dummy_cycles = dummy_cycles;
			}
		}


>  		reinit_completion(&tqspi->xfer_completion);
>  
>  		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);

[PATCH v4 7/9] arm64: tegra: Enable QSPI on Jetson Nano

[Sowjanya Komatineni]

This patch enables QSPI on Jetson Nano.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts | 12 ++++++++++++
 arch/arm64/boot/dts/nvidia/tegra210.dtsi           |  5 +++--
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts b/arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts
index 6a877de..a1b4603 100644
--- a/arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts
+++ b/arch/arm64/boot/dts/nvidia/tegra210-p3450-0000.dts
@@ -638,6 +638,18 @@
 		};
 	};
 
+	spi@70410000 {
+		status = "okay";
+
+		flash@0 {
+			compatible = "spi-nor";
+			reg = <0>;
+			spi-max-frequency = <104000000>;
+			spi-tx-bus-width = <2>;
+			spi-rx-bus-width = <2>;
+		};
+	};
+
 	clk32k_in: clock@0 {
 		compatible = "fixed-clock";
 		clock-frequency = <32768>;
diff --git a/arch/arm64/boot/dts/nvidia/tegra210.dtsi b/arch/arm64/boot/dts/nvidia/tegra210.dtsi
index 4fbf8c1..998fa81 100644
--- a/arch/arm64/boot/dts/nvidia/tegra210.dtsi
+++ b/arch/arm64/boot/dts/nvidia/tegra210.dtsi
@@ -1536,8 +1536,9 @@
 		interrupts = <GIC_SPI 10 IRQ_TYPE_LEVEL_HIGH>;
 		#address-cells = <1>;
 		#size-cells = <0>;
-		clocks = <&tegra_car TEGRA210_CLK_QSPI>;
-		clock-names = "qspi";
+		clocks = <&tegra_car TEGRA210_CLK_QSPI>,
+			 <&tegra_car TEGRA210_CLK_QSPI_PM>;
+		clock-names = "qspi", "qspi_out";
 		resets = <&tegra_car 211>;
 		reset-names = "qspi";
 		dmas = <&apbdma 5>, <&apbdma 5>;
-- 
2.7.4

[PATCH v4 8/9] arm64: tegra: Add QSPI nodes on Tegra194

[Sowjanya Komatineni]

Tegra194 has 2 QSPI controllers.

This patch adds DT node for these 2 QSPI controllers.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 arch/arm64/boot/dts/nvidia/tegra194.dtsi | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/arch/arm64/boot/dts/nvidia/tegra194.dtsi b/arch/arm64/boot/dts/nvidia/tegra194.dtsi
index 25f36d6..852980f 100644
--- a/arch/arm64/boot/dts/nvidia/tegra194.dtsi
+++ b/arch/arm64/boot/dts/nvidia/tegra194.dtsi
@@ -609,6 +609,34 @@
 			status = "disabled";
 		};
 
+		spi@3270000 {
+			compatible = "nvidia,tegra194-qspi";
+			reg = <0x3270000 0x1000>;
+			interrupts = <GIC_SPI 35 IRQ_TYPE_LEVEL_HIGH>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			clocks = <&bpmp TEGRA194_CLK_QSPI0>,
+				 <&bpmp TEGRA194_CLK_QSPI0_PM>;
+			clock-names = "qspi", "qspi_out";
+			resets = <&bpmp TEGRA194_RESET_QSPI0>;
+			reset-names = "qspi";
+			status = "disabled";
+		};
+
+		spi@3300000 {
+			compatible = "nvidia,tegra194-qspi";
+			reg = <0x3300000 0x1000>;
+			interrupts = <GIC_SPI 39 IRQ_TYPE_LEVEL_HIGH>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			clocks = <&bpmp TEGRA194_CLK_QSPI1>,
+				 <&bpmp TEGRA194_CLK_QSPI1_PM>;
+			clock-names = "qspi", "qspi_out";
+			resets = <&bpmp TEGRA194_RESET_QSPI1>;
+			reset-names = "qspi";
+			status = "disabled";
+		};
+
 		pwm1: pwm@3280000 {
 			compatible = "nvidia,tegra194-pwm",
 				     "nvidia,tegra186-pwm";
-- 
2.7.4

[PATCH v4 9/9] arm64: tegra: Enable QSPI on Jetson Xavier NX

[Sowjanya Komatineni]

This patch enables QSPI on Jetson Xavier NX.

Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
---
 .../arm64/boot/dts/nvidia/tegra194-p3509-0000+p3668-0000.dts | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/arch/arm64/boot/dts/nvidia/tegra194-p3509-0000+p3668-0000.dts b/arch/arm64/boot/dts/nvidia/tegra194-p3509-0000+p3668-0000.dts
index 7f97b34..f1053e7 100644
--- a/arch/arm64/boot/dts/nvidia/tegra194-p3509-0000+p3668-0000.dts
+++ b/arch/arm64/boot/dts/nvidia/tegra194-p3509-0000+p3668-0000.dts
@@ -100,6 +100,18 @@
 			phy-names = "usb2-1", "usb2-2", "usb3-2";
 		};
 
+		spi@3270000 {
+			status = "okay";
+
+			flash@0 {
+				compatible = "spi-nor";
+				reg = <0>;
+				spi-max-frequency = <102000000>;
+				spi-tx-bus-width = <4>;
+				spi-rx-bus-width = <4>;
+			};
+		};
+
 		pwm@32d0000 {
 			status = "okay";
 		};
-- 
2.7.4