[PATCH v9 0/9] Marvell HW overlay support for Cadence xSPI

[PATCH v9 0/9] Marvell HW overlay support for Cadence xSPI

[Witold Sadowski]

This patch series adds support for the second version of the Marvell
hardware overlay for the Cadence xSPI IP block. The overlay is a hardware
change made around the original xSPI block. It extends xSPI features with
clock configuration, interrupt masking, and full-duplex, variable-length SPI
operations.

These functionalities allow the xSPI block to operate not only with memory
devices but also with simple SPI devices and TPM devices.

Example ACPI entry:
      Device (SPI0) {
        Name (_HID, "PRP0001")          // ACPI_DT_NAMESPACE_HID
        Name (_UID, 0)
        Name (_DDN, "SPI controller 0")
        Name (_CCA, ONE)

        Method (_STA) {Return (0xF)}

        Name (_CRS, ResourceTemplate() {

          QWordMemory ( ResourceConsumer,// ResourceUsage
                        PosDecode,       // Decode
                        MinFixed,        // MinType
                        MaxFixed,        // MaxType
                        NonCacheable,    // MemType
                        ReadWrite,       // ReadWriteType
                        0,               // AddressGranularity
                        0x804000000000,  // MinAddress
                        0x804000001037,  // MaxAddress
                        0,               // AddressTranslation
                        0x1038)          // AddressLength

          QWordMemory ( ResourceConsumer,// ResourceUsage
                        PosDecode,       // Decode
                        MinFixed,        // MinType
                        MaxFixed,        // MaxType
                        NonCacheable,    // MemType
                        ReadWrite,       // ReadWriteType
                        0,               // AddressGranularity
                        0x804010000000,  // MinAddress
                        0x804010000007,  // MaxAddress
                        0,               // AddressTranslation
                        0x8)             // AddressLength

          QWordMemory ( ResourceConsumer,// ResourceUsage
                        PosDecode,       // Decode
                        MinFixed,        // MinType
                        MaxFixed,        // MaxType
                        NonCacheable,    // MemType
                        ReadWrite,       // ReadWriteType
                        0,               // AddressGranularity
                        0x804000002000,  // MinAddress
                        0x804000004027,  // MaxAddress
                        0,               // AddressTranslation
                        0x2028)          // AddressLength

          QWordMemory ( ResourceConsumer,// ResourceUsage
            PosDecode,       // Decode
            MinFixed,        // MinType
            MaxFixed,        // MaxType
            NonCacheable,    // MemType
            ReadWrite,       // ReadWriteType
            0,               // AddressGranularity
            0x804000008000,  // MinAddress
            0x804000008237,  // MaxAddress
            0,               // AddressTranslation
            0x238)           // AddressLength

          Interrupt(ResourceConsumer, Edge, ActiveHigh, Exclusive) { 0x7A }
        })

        Name (_DSD, Package() {
            ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"),
            Package () {
                Package () { "compatible", "marvell,cn10-xspi-nor"},
                Package () { "reg", 0x8040},
            }
        })
      } // SPI0

Changes:
v9:
  Split into smaller patches:
    - Marvell overlay splitted into: PHY, Clock, Interrupt and SDMA ops
    - ACPI support splitted into resource mapping, CS parameter reading and tx/rx bus length
  Add separate ops and a few function pointers to distinguish between Cadence and Marvell:
    - SDMA read handler.
    - IRQ enable/disable handler
    - Separate mem_ops for Marvell xSPI
  Cleanup xfer code from magic numbers
  Add more descriptive commit msg for xfer block
  Use bitrev8 instead of custom bit reversal
  Rework Marvell SDMA read operations
  Add example ACPI entry

v8:
  Rename xferbase to xfer
  Rework DLL reset, to return non inverted boolean value
  Rework STIG and SDMA status check, to return non inverted boolean value

v7:
  Rebase patches to latest sources, changes in "Allow to read basic xSPI configuration
 from ACPI"
  Removed bugfix, as it was integrated to next tree from v6

v6:
  Fix item order in cdns,xspi.yaml

v5:
  Rework cdns,xspi.yaml file
  Reword commit messages
  Move mamory mapping to ACPI patch
  Use devm_platform_ioremap_resource instead of two step mapping

v4:
  Rename new Marvell registers to keep naming conventions
  Rename mrvl,xspi-nor to marvell,cnxx,xspi-nor
  Various fixed for cdns,xspi.yaml file:
    - Remove unnecesary parameters
    - Link register xferbase with marvell,cn10-xspi-nor
    - Move default values to .c file from device-tree
  Clock configuration optimization
  ACPI fixes:
    - Remove incorrect ACPI match table
  Added .data field to device_id, fixes for matching in ACPI and dtb case
  Minor style comment changes

v3:
  Removed all kconfig changes
  Added device-tree mrvl,xspi-nor tag

v2:
  Support for second overlay iteration

v1:
  -

v0:
  Initial support for v1 overlay

Witold Sadowski (9):
  spi: dt-bindings: cadence: Add Marvell overlay bindings documentation
    for Cadence XSPI
  spi: cadence: Add static PHY configuration in Marvell overlay
  spi: cadence: Add clock configuration for Marvell xSPI overlay
  spi: cadence: Add Marvell SDMA operations
  spi: cadence: Add Marvell xSPI interrupt changes
  spi: cadence: Add Marvell xfer operation support
  spi: cadence: Change resource mapping
  spi: cadence: Change cs property reading.
  spi: cadence: Try to read spi-tx/rx-bus width property using ACPI

 .../devicetree/bindings/spi/cdns,xspi.yaml    |  32 +-
 drivers/spi/spi-cadence-xspi.c                | 665 +++++++++++++++++-
 2 files changed, 677 insertions(+), 20 deletions(-)

-- 
2.43.0

[PATCH v9 1/9] spi: dt-bindings: cadence: Add Marvell overlay bindings documentation for Cadence XSPI

[Witold Sadowski]

Add new bindings for the v2 Marvell xSPI overlay: marvell,cn10-xspi-nor
compatible string. This new compatible string distinguishes between the
original and modified xSPI block.

Also add an optional base for the xfer register set with an additional
reg field to allocate the xSPI Marvell overlay XFER block.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
---
 .../devicetree/bindings/spi/cdns,xspi.yaml    | 32 ++++++++++++++++---
 1 file changed, 28 insertions(+), 4 deletions(-)

diff --git a/Documentation/devicetree/bindings/spi/cdns,xspi.yaml b/Documentation/devicetree/bindings/spi/cdns,xspi.yaml
index eb0f92468185..38a5795589de 100644
--- a/Documentation/devicetree/bindings/spi/cdns,xspi.yaml
+++ b/Documentation/devicetree/bindings/spi/cdns,xspi.yaml
@@ -15,24 +15,27 @@ description: |
   single, dual, quad or octal wire transmission modes for
   read/write access to slaves such as SPI-NOR flash.
 
-allOf:
-  - $ref: spi-controller.yaml#
-
 properties:
   compatible:
-    const: cdns,xspi-nor
+    enum:
+      - cdns,xspi-nor
+      - marvell,cn10-xspi-nor
 
   reg:
     items:
       - description: address and length of the controller register set
       - description: address and length of the Slave DMA data port
       - description: address and length of the auxiliary registers
+      - description: address and length of the xfer registers
+    minItems: 3
 
   reg-names:
     items:
       - const: io
       - const: sdma
       - const: aux
+      - const: xfer
+    minItems: 3
 
   interrupts:
     maxItems: 1
@@ -42,6 +45,27 @@ required:
   - reg
   - interrupts
 
+allOf:
+  - $ref: spi-controller.yaml#
+  - if:
+      properties:
+        compatible:
+          contains:
+            enum:
+              - marvell,cn10-xspi-nor
+    then:
+      properties:
+        reg:
+          minItems: 4
+        reg-names:
+          minItems: 4
+    else:
+      properties:
+        reg:
+          maxItems: 3
+        reg-names:
+          maxItems: 3
+
 unevaluatedProperties: false
 
 examples:
-- 
2.43.0

[PATCH v9 2/9] spi: cadence: Add static PHY configuration in Marvell overlay

[Witold Sadowski]

This commit adds support for static PHY configuration of Cadence xSPI
block. Configuration will be applied only if Marvell overlay compatible
string will be detected. Configuration is static over the whole
frequency range.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 113 ++++++++++++++++++++++++++++++++-
 1 file changed, 112 insertions(+), 1 deletion(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index 2e3eacd46b72..70b1b4a0ff13 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -193,6 +193,30 @@
 		((op)->data.dir == SPI_MEM_DATA_IN) ? \
 		CDNS_XSPI_STIG_CMD_DIR_READ : CDNS_XSPI_STIG_CMD_DIR_WRITE))
 
+/* Marvell PHY default values */
+#define MARVELL_REGS_DLL_PHY_CTRL		0x00000707
+#define MARVELL_CTB_RFILE_PHY_CTRL		0x00004000
+#define MARVELL_RFILE_PHY_TSEL			0x00000000
+#define MARVELL_RFILE_PHY_DQ_TIMING		0x00000101
+#define MARVELL_RFILE_PHY_DQS_TIMING		0x00700404
+#define MARVELL_RFILE_PHY_GATE_LPBK_CTRL	0x00200030
+#define MARVELL_RFILE_PHY_DLL_MASTER_CTRL	0x00800000
+#define MARVELL_RFILE_PHY_DLL_SLAVE_CTRL	0x0000ff01
+
+/* PHY config registers */
+#define CDNS_XSPI_RF_MINICTRL_REGS_DLL_PHY_CTRL			0x1034
+#define CDNS_XSPI_PHY_CTB_RFILE_PHY_CTRL			0x0080
+#define CDNS_XSPI_PHY_CTB_RFILE_PHY_TSEL			0x0084
+#define CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DQ_TIMING		0x0000
+#define CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DQS_TIMING		0x0004
+#define CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_GATE_LPBK_CTRL	0x0008
+#define CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DLL_MASTER_CTRL	0x000c
+#define CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DLL_SLAVE_CTRL	0x0010
+#define CDNS_XSPI_DATASLICE_RFILE_PHY_DLL_OBS_REG_0		0x001c
+
+#define CDNS_XSPI_DLL_RST_N BIT(24)
+#define CDNS_XSPI_DLL_LOCK  BIT(0)
+
 enum cdns_xspi_stig_instr_type {
 	CDNS_XSPI_STIG_INSTR_TYPE_0,
 	CDNS_XSPI_STIG_INSTR_TYPE_1,
@@ -209,6 +233,34 @@ enum cdns_xspi_stig_cmd_dir {
 	CDNS_XSPI_STIG_CMD_DIR_WRITE,
 };
 
+struct cdns_xspi_driver_data {
+	bool mrvl_hw_overlay;
+	u32 dll_phy_ctrl;
+	u32 ctb_rfile_phy_ctrl;
+	u32 rfile_phy_tsel;
+	u32 rfile_phy_dq_timing;
+	u32 rfile_phy_dqs_timing;
+	u32 rfile_phy_gate_lpbk_ctrl;
+	u32 rfile_phy_dll_master_ctrl;
+	u32 rfile_phy_dll_slave_ctrl;
+};
+
+static struct cdns_xspi_driver_data marvell_driver_data = {
+	.mrvl_hw_overlay = true,
+	.dll_phy_ctrl = MARVELL_REGS_DLL_PHY_CTRL,
+	.ctb_rfile_phy_ctrl = MARVELL_CTB_RFILE_PHY_CTRL,
+	.rfile_phy_tsel = MARVELL_RFILE_PHY_TSEL,
+	.rfile_phy_dq_timing = MARVELL_RFILE_PHY_DQ_TIMING,
+	.rfile_phy_dqs_timing = MARVELL_RFILE_PHY_DQS_TIMING,
+	.rfile_phy_gate_lpbk_ctrl = MARVELL_RFILE_PHY_GATE_LPBK_CTRL,
+	.rfile_phy_dll_master_ctrl = MARVELL_RFILE_PHY_DLL_MASTER_CTRL,
+	.rfile_phy_dll_slave_ctrl = MARVELL_RFILE_PHY_DLL_SLAVE_CTRL,
+};
+
+static struct cdns_xspi_driver_data cdns_driver_data = {
+	.mrvl_hw_overlay = false,
+};
+
 struct cdns_xspi_dev {
 	struct platform_device *pdev;
 	struct device *dev;
@@ -230,8 +282,55 @@ struct cdns_xspi_dev {
 	const void *out_buffer;
 
 	u8 hw_num_banks;
+
+	const struct cdns_xspi_driver_data *driver_data;
 };
 
+static void cdns_xspi_reset_dll(struct cdns_xspi_dev *cdns_xspi)
+{
+	u32 dll_cntrl = readl(cdns_xspi->iobase +
+			      CDNS_XSPI_RF_MINICTRL_REGS_DLL_PHY_CTRL);
+
+	/* Reset DLL */
+	dll_cntrl |= CDNS_XSPI_DLL_RST_N;
+	writel(dll_cntrl, cdns_xspi->iobase +
+			  CDNS_XSPI_RF_MINICTRL_REGS_DLL_PHY_CTRL);
+}
+
+static bool cdns_xspi_is_dll_locked(struct cdns_xspi_dev *cdns_xspi)
+{
+	u32 dll_lock;
+
+	return !readl_relaxed_poll_timeout(cdns_xspi->iobase +
+		CDNS_XSPI_INTR_STATUS_REG,
+		dll_lock, ((dll_lock & CDNS_XSPI_DLL_LOCK) == 1), 10, 10000);
+}
+
+/* Static configuration of PHY */
+static bool cdns_xspi_configure_phy(struct cdns_xspi_dev *cdns_xspi)
+{
+	writel(cdns_xspi->driver_data->dll_phy_ctrl,
+	       cdns_xspi->iobase + CDNS_XSPI_RF_MINICTRL_REGS_DLL_PHY_CTRL);
+	writel(cdns_xspi->driver_data->ctb_rfile_phy_ctrl,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_CTB_RFILE_PHY_CTRL);
+	writel(cdns_xspi->driver_data->rfile_phy_tsel,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_CTB_RFILE_PHY_TSEL);
+	writel(cdns_xspi->driver_data->rfile_phy_dq_timing,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DQ_TIMING);
+	writel(cdns_xspi->driver_data->rfile_phy_dqs_timing,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DQS_TIMING);
+	writel(cdns_xspi->driver_data->rfile_phy_gate_lpbk_ctrl,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_GATE_LPBK_CTRL);
+	writel(cdns_xspi->driver_data->rfile_phy_dll_master_ctrl,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DLL_MASTER_CTRL);
+	writel(cdns_xspi->driver_data->rfile_phy_dll_slave_ctrl,
+	       cdns_xspi->auxbase + CDNS_XSPI_PHY_DATASLICE_RFILE_PHY_DLL_SLAVE_CTRL);
+
+	cdns_xspi_reset_dll(cdns_xspi);
+
+	return cdns_xspi_is_dll_locked(cdns_xspi);
+}
+
 static int cdns_xspi_wait_for_controller_idle(struct cdns_xspi_dev *cdns_xspi)
 {
 	u32 ctrl_stat;
@@ -544,13 +643,17 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 		SPI_RX_DUAL | SPI_RX_QUAD | SPI_TX_OCTAL | SPI_RX_OCTAL |
 		SPI_MODE_0  | SPI_MODE_3;
 
+	cdns_xspi = spi_controller_get_devdata(host);
+	cdns_xspi->driver_data = of_device_get_match_data(dev);
+	if (!cdns_xspi->driver_data)
+		return -ENODEV;
+
 	host->mem_ops = &cadence_xspi_mem_ops;
 	host->dev.of_node = pdev->dev.of_node;
 	host->bus_num = -1;
 
 	platform_set_drvdata(pdev, host);
 
-	cdns_xspi = spi_controller_get_devdata(host);
 	cdns_xspi->pdev = pdev;
 	cdns_xspi->dev = &pdev->dev;
 	cdns_xspi->cur_cs = 0;
@@ -592,6 +695,9 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 		return ret;
 	}
 
+	if (cdns_xspi->driver_data->mrvl_hw_overlay)
+		cdns_xspi_configure_phy(cdns_xspi);
+
 	cdns_xspi_print_phy_config(cdns_xspi);
 
 	ret = cdns_xspi_controller_init(cdns_xspi);
@@ -616,6 +722,11 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 static const struct of_device_id cdns_xspi_of_match[] = {
 	{
 		.compatible = "cdns,xspi-nor",
+		.data = &cdns_driver_data,
+	},
+	{
+		.compatible = "marvell,cn10-xspi-nor",
+		.data = &marvell_driver_data,
 	},
 	{ /* end of table */}
 };
-- 
2.43.0

[PATCH v9 3/9] spi: cadence: Add clock configuration for Marvell xSPI overlay

[Witold Sadowski]

Add support for clock divider. Divider block can disable, enable and
divide clock signal. Only 14 different divide ratios are avalible, from
6.25 up to 200MHz. For calculations use default Marvell system clock
value(800MHz).

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 97 +++++++++++++++++++++++++++++++++-
 1 file changed, 95 insertions(+), 2 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index 70b1b4a0ff13..d0222284c507 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -217,6 +217,15 @@
 #define CDNS_XSPI_DLL_RST_N BIT(24)
 #define CDNS_XSPI_DLL_LOCK  BIT(0)
 
+/* Marvell overlay registers - clock */
+#define MRVL_XSPI_CLK_CTRL_AUX_REG   0x2020
+#define MRVL_XSPI_CLK_ENABLE	     BIT(0)
+#define MRVL_XSPI_CLK_DIV	     GENMASK(4, 1)
+#define MRVL_XSPI_IRQ_ENABLE	     BIT(6)
+#define MRVL_XSPI_CLOCK_IO_HZ	     800000000
+#define MRVL_XSPI_CLOCK_DIVIDED(div) ((MRVL_XSPI_CLOCK_IO_HZ) / (div))
+#define MRVL_DEFAULT_CLK	     25000000
+
 enum cdns_xspi_stig_instr_type {
 	CDNS_XSPI_STIG_INSTR_TYPE_0,
 	CDNS_XSPI_STIG_INSTR_TYPE_1,
@@ -261,6 +270,23 @@ static struct cdns_xspi_driver_data cdns_driver_data = {
 	.mrvl_hw_overlay = false,
 };
 
+const int cdns_mrvl_xspi_clk_div_list[] = {
+	4,	//0x0 = Divide by 4.   SPI clock is 200 MHz.
+	6,	//0x1 = Divide by 6.   SPI clock is 133.33 MHz.
+	8,	//0x2 = Divide by 8.   SPI clock is 100 MHz.
+	10,	//0x3 = Divide by 10.  SPI clock is 80 MHz.
+	12,	//0x4 = Divide by 12.  SPI clock is 66.666 MHz.
+	16,	//0x5 = Divide by 16.  SPI clock is 50 MHz.
+	18,	//0x6 = Divide by 18.  SPI clock is 44.44 MHz.
+	20,	//0x7 = Divide by 20.  SPI clock is 40 MHz.
+	24,	//0x8 = Divide by 24.  SPI clock is 33.33 MHz.
+	32,	//0x9 = Divide by 32.  SPI clock is 25 MHz.
+	40,	//0xA = Divide by 40.  SPI clock is 20 MHz.
+	50,	//0xB = Divide by 50.  SPI clock is 16 MHz.
+	64,	//0xC = Divide by 64.  SPI clock is 12.5 MHz.
+	128	//0xD = Divide by 128. SPI clock is 6.25 MHz.
+};
+
 struct cdns_xspi_dev {
 	struct platform_device *pdev;
 	struct device *dev;
@@ -331,6 +357,48 @@ static bool cdns_xspi_configure_phy(struct cdns_xspi_dev *cdns_xspi)
 	return cdns_xspi_is_dll_locked(cdns_xspi);
 }
 
+static bool cdns_mrvl_xspi_setup_clock(struct cdns_xspi_dev *cdns_xspi,
+				       int requested_clk)
+{
+	int i = 0;
+	int clk_val;
+	u32 clk_reg;
+	bool update_clk = false;
+
+	while (i < ARRAY_SIZE(cdns_mrvl_xspi_clk_div_list)) {
+		clk_val = MRVL_XSPI_CLOCK_DIVIDED(
+				cdns_mrvl_xspi_clk_div_list[i]);
+		if (clk_val <= requested_clk)
+			break;
+		i++;
+	}
+
+	dev_dbg(cdns_xspi->dev, "Found clk div: %d, clk val: %d\n",
+		cdns_mrvl_xspi_clk_div_list[i],
+		MRVL_XSPI_CLOCK_DIVIDED(
+		cdns_mrvl_xspi_clk_div_list[i]));
+
+	clk_reg = readl(cdns_xspi->auxbase + MRVL_XSPI_CLK_CTRL_AUX_REG);
+
+	if (FIELD_GET(MRVL_XSPI_CLK_DIV, clk_reg) != i) {
+		clk_reg &= ~MRVL_XSPI_CLK_ENABLE;
+		writel(clk_reg,
+		       cdns_xspi->auxbase + MRVL_XSPI_CLK_CTRL_AUX_REG);
+		clk_reg = FIELD_PREP(MRVL_XSPI_CLK_DIV, i);
+		clk_reg &= ~MRVL_XSPI_CLK_DIV;
+		clk_reg |= FIELD_PREP(MRVL_XSPI_CLK_DIV, i);
+		clk_reg |= MRVL_XSPI_CLK_ENABLE;
+		clk_reg |= MRVL_XSPI_IRQ_ENABLE;
+		update_clk = true;
+	}
+
+	if (update_clk)
+		writel(clk_reg,
+		       cdns_xspi->auxbase + MRVL_XSPI_CLK_CTRL_AUX_REG);
+
+	return update_clk;
+}
+
 static int cdns_xspi_wait_for_controller_idle(struct cdns_xspi_dev *cdns_xspi)
 {
 	u32 ctrl_stat;
@@ -536,6 +604,21 @@ static int cdns_xspi_mem_op_execute(struct spi_mem *mem,
 	return ret;
 }
 
+static int marvell_xspi_mem_op_execute(struct spi_mem *mem,
+				    const struct spi_mem_op *op)
+{
+	struct cdns_xspi_dev *cdns_xspi =
+		spi_controller_get_devdata(mem->spi->controller);
+	int ret = 0;
+
+	cdns_mrvl_xspi_setup_clock(cdns_xspi, mem->spi->max_speed_hz);
+
+	ret = cdns_xspi_mem_op(cdns_xspi, mem, op);
+
+	return ret;
+}
+
+
 static int cdns_xspi_adjust_mem_op_size(struct spi_mem *mem, struct spi_mem_op *op)
 {
 	struct cdns_xspi_dev *cdns_xspi =
@@ -551,6 +634,11 @@ static const struct spi_controller_mem_ops cadence_xspi_mem_ops = {
 	.adjust_op_size = cdns_xspi_adjust_mem_op_size,
 };
 
+static const struct spi_controller_mem_ops marvell_xspi_mem_ops = {
+	.exec_op = marvell_xspi_mem_op_execute,
+	.adjust_op_size = cdns_xspi_adjust_mem_op_size,
+};
+
 static irqreturn_t cdns_xspi_irq_handler(int this_irq, void *dev)
 {
 	struct cdns_xspi_dev *cdns_xspi = dev;
@@ -648,7 +736,10 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 	if (!cdns_xspi->driver_data)
 		return -ENODEV;
 
-	host->mem_ops = &cadence_xspi_mem_ops;
+	if (cdns_xspi->driver_data->mrvl_hw_overlay)
+		host->mem_ops = &marvell_xspi_mem_ops;
+	else
+		host->mem_ops = &cadence_xspi_mem_ops;
 	host->dev.of_node = pdev->dev.of_node;
 	host->bus_num = -1;
 
@@ -695,8 +786,10 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 		return ret;
 	}
 
-	if (cdns_xspi->driver_data->mrvl_hw_overlay)
+	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
+		cdns_mrvl_xspi_setup_clock(cdns_xspi, MRVL_DEFAULT_CLK);
 		cdns_xspi_configure_phy(cdns_xspi);
+	}
 
 	cdns_xspi_print_phy_config(cdns_xspi);
 
-- 
2.43.0

[PATCH v9 4/9] spi: cadence: Add Marvell SDMA operations

[Witold Sadowski]

In Marvell xSPI implementation any access to SDMA register will result
in 8 byte SPI data transfer. Reading less data(eg. 1B) will result in
losing remaining bytes. To avoid that read/write 8 bytes into temporary
buffer, and read/write whole temporary buffer into SDMA.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 56 ++++++++++++++++++++++++++++++++--
 1 file changed, 53 insertions(+), 3 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index d0222284c507..c79f2a2931a8 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -310,6 +310,7 @@ struct cdns_xspi_dev {
 	u8 hw_num_banks;
 
 	const struct cdns_xspi_driver_data *driver_data;
+	void (*sdma_handler)(struct cdns_xspi_dev *cdns_xspi);
 };
 
 static void cdns_xspi_reset_dll(struct cdns_xspi_dev *cdns_xspi)
@@ -515,6 +516,52 @@ static void cdns_xspi_sdma_handle(struct cdns_xspi_dev *cdns_xspi)
 	}
 }
 
+static void m_ioreadq(void __iomem  *addr, void *buf, int len)
+{
+	u64 tmp_buf;
+
+	while (len) {
+		tmp_buf = readq(addr);
+		memcpy(buf, &tmp_buf, len > 8 ? 8 : len);
+		len = len > 8 ? len - 8 : 0;
+		buf += 8;
+	}
+}
+
+static void m_iowriteq(void __iomem *addr, const void *buf, int len)
+{
+	u64 tmp_buf;
+
+	while (len) {
+		memcpy(&tmp_buf, buf, len > 8 ? 8 : len);
+		writeq(tmp_buf, addr);
+		len = len > 8 ? len - 8 : 0;
+		buf += 8;
+	}
+}
+
+static void marvell_xspi_sdma_handle(struct cdns_xspi_dev *cdns_xspi)
+{
+	u32 sdma_size, sdma_trd_info;
+	u8 sdma_dir;
+
+	sdma_size = readl(cdns_xspi->iobase + CDNS_XSPI_SDMA_SIZE_REG);
+	sdma_trd_info = readl(cdns_xspi->iobase + CDNS_XSPI_SDMA_TRD_INFO_REG);
+	sdma_dir = FIELD_GET(CDNS_XSPI_SDMA_DIR, sdma_trd_info);
+
+	switch (sdma_dir) {
+	case CDNS_XSPI_SDMA_DIR_READ:
+		m_ioreadq(cdns_xspi->sdmabase,
+			    cdns_xspi->in_buffer, sdma_size);
+		break;
+
+	case CDNS_XSPI_SDMA_DIR_WRITE:
+		m_iowriteq(cdns_xspi->sdmabase,
+			     cdns_xspi->out_buffer, sdma_size);
+		break;
+	}
+}
+
 static int cdns_xspi_send_stig_command(struct cdns_xspi_dev *cdns_xspi,
 				       const struct spi_mem_op *op,
 				       bool data_phase)
@@ -566,7 +613,7 @@ static int cdns_xspi_send_stig_command(struct cdns_xspi_dev *cdns_xspi,
 			cdns_xspi_set_interrupts(cdns_xspi, false);
 			return -EIO;
 		}
-		cdns_xspi_sdma_handle(cdns_xspi);
+		cdns_xspi->sdma_handler(cdns_xspi);
 	}
 
 	wait_for_completion(&cdns_xspi->cmd_complete);
@@ -736,10 +783,13 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 	if (!cdns_xspi->driver_data)
 		return -ENODEV;
 
-	if (cdns_xspi->driver_data->mrvl_hw_overlay)
+	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
 		host->mem_ops = &marvell_xspi_mem_ops;
-	else
+		cdns_xspi->sdma_handler = &marvell_xspi_sdma_handle;
+	} else {
 		host->mem_ops = &cadence_xspi_mem_ops;
+		cdns_xspi->sdma_handler = &cdns_xspi_sdma_handle;
+	}
 	host->dev.of_node = pdev->dev.of_node;
 	host->bus_num = -1;
 
-- 
2.43.0

Re: [PATCH v9 4/9] spi: cadence: Add Marvell SDMA operations

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 448 bytes --]

On Wed, Jun 19, 2024 at 07:17:10AM -0700, Witold Sadowski wrote:

> +static void m_ioreadq(void __iomem  *addr, void *buf, int len)
> +{
> +	u64 tmp_buf;
> +
> +	while (len) {
> +		tmp_buf = readq(addr);
> +		memcpy(buf, &tmp_buf, len > 8 ? 8 : len);
> +		len = len > 8 ? len - 8 : 0;
> +		buf += 8;
> +	}
> +}

Wouldn't it be more efficient and readable to only do the memcpy() for
the trailing bytes and just do this memcpy() for the final word?

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

RE: [EXTERNAL] Re: [PATCH v9 4/9] spi: cadence: Add Marvell SDMA operations

[Witold Sadowski]

Hi Mark
> 
> > +static void m_ioreadq(void __iomem  *addr, void *buf, int len) {
> > +	u64 tmp_buf;
> > +
> > +	while (len) {
> > +		tmp_buf = readq(addr);
> > +		memcpy(buf, &tmp_buf, len > 8 ? 8 : len);
> > +		len = len > 8 ? len - 8 : 0;
> > +		buf += 8;
> > +	}
> > +}
> 
> Wouldn't it be more efficient and readable to only do the memcpy() for the
> trailing bytes and just do this memcpy() for the final word?

The whole problem is with SDMA end - addr in that case. If code will try to
Read it in non 64b mode, all remaining bits, will be lost.
For example - doing 1B read on that register, will return 1B, but SDMA will
transfer 8B, dropping remaining 7B.
I have tried memcpy approach, and it was not stable.

Regards
Witek 

Re: [EXTERNAL] Re: [PATCH v9 4/9] spi: cadence: Add Marvell SDMA operations

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 828 bytes --]

On Fri, Jun 28, 2024 at 01:45:13PM +0000, Witold Sadowski wrote:
> Hi Mark

> > > +	while (len) {
> > > +		tmp_buf = readq(addr);
> > > +		memcpy(buf, &tmp_buf, len > 8 ? 8 : len);
> > > +		len = len > 8 ? len - 8 : 0;
> > > +		buf += 8;
> > > +	}
> > > +}

> > Wouldn't it be more efficient and readable to only do the memcpy() for the
> > trailing bytes and just do this memcpy() for the final word?

> The whole problem is with SDMA end - addr in that case. If code will try to
> Read it in non 64b mode, all remaining bits, will be lost.
> For example - doing 1B read on that register, will return 1B, but SDMA will
> transfer 8B, dropping remaining 7B.
> I have tried memcpy approach, and it was not stable.

That's not what I'm suggesting, like I said above I'm suggesting to
*only* do the memcpy() for the trailing word.

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

[PATCH v9 5/9] spi: cadence: Add Marvell xSPI interrupt changes

[Witold Sadowski]

It is possible that before enabling interrupt, interrupt bit will be
set. It might cause improper IRQ handler behaviour. To fix it, clear
interrupt bit before enabling interrupts. That behaviour is specific to
Marvell xSPI implementation.
In addition in Marvell xSPI interrupt must be cleared in two places -
xSPI itself, and Marvell overlay.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 28 ++++++++++++++++++++++++----
 1 file changed, 24 insertions(+), 4 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index c79f2a2931a8..eb9293e8d708 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -311,6 +311,7 @@ struct cdns_xspi_dev {
 
 	const struct cdns_xspi_driver_data *driver_data;
 	void (*sdma_handler)(struct cdns_xspi_dev *cdns_xspi);
+	void (*set_interrupts_handler)(struct cdns_xspi_dev *cdns_xspi, bool enabled);
 };
 
 static void cdns_xspi_reset_dll(struct cdns_xspi_dev *cdns_xspi)
@@ -472,6 +473,23 @@ static void cdns_xspi_set_interrupts(struct cdns_xspi_dev *cdns_xspi,
 	writel(intr_enable, cdns_xspi->iobase + CDNS_XSPI_INTR_ENABLE_REG);
 }
 
+static void marvell_xspi_set_interrupts(struct cdns_xspi_dev *cdns_xspi,
+				     bool enabled)
+{
+	u32 intr_enable;
+	u32 irq_status;
+
+	irq_status = readl(cdns_xspi->iobase + CDNS_XSPI_INTR_STATUS_REG);
+	writel(irq_status, cdns_xspi->iobase + CDNS_XSPI_INTR_STATUS_REG);
+
+	intr_enable = readl(cdns_xspi->iobase + CDNS_XSPI_INTR_ENABLE_REG);
+	if (enabled)
+		intr_enable |= CDNS_XSPI_INTR_MASK;
+	else
+		intr_enable &= ~CDNS_XSPI_INTR_MASK;
+	writel(intr_enable, cdns_xspi->iobase + CDNS_XSPI_INTR_ENABLE_REG);
+}
+
 static int cdns_xspi_controller_init(struct cdns_xspi_dev *cdns_xspi)
 {
 	u32 ctrl_ver;
@@ -489,7 +507,7 @@ static int cdns_xspi_controller_init(struct cdns_xspi_dev *cdns_xspi)
 
 	ctrl_features = readl(cdns_xspi->iobase + CDNS_XSPI_CTRL_FEATURES_REG);
 	cdns_xspi->hw_num_banks = FIELD_GET(CDNS_XSPI_NUM_BANKS, ctrl_features);
-	cdns_xspi_set_interrupts(cdns_xspi, false);
+	cdns_xspi->set_interrupts_handler(cdns_xspi, false);
 
 	return 0;
 }
@@ -578,7 +596,7 @@ static int cdns_xspi_send_stig_command(struct cdns_xspi_dev *cdns_xspi,
 	writel(FIELD_PREP(CDNS_XSPI_CTRL_WORK_MODE, CDNS_XSPI_WORK_MODE_STIG),
 	       cdns_xspi->iobase + CDNS_XSPI_CTRL_CONFIG_REG);
 
-	cdns_xspi_set_interrupts(cdns_xspi, true);
+	cdns_xspi->set_interrupts_handler(cdns_xspi, true);
 	cdns_xspi->sdma_error = false;
 
 	memset(cmd_regs, 0, sizeof(cmd_regs));
@@ -610,14 +628,14 @@ static int cdns_xspi_send_stig_command(struct cdns_xspi_dev *cdns_xspi,
 
 		wait_for_completion(&cdns_xspi->sdma_complete);
 		if (cdns_xspi->sdma_error) {
-			cdns_xspi_set_interrupts(cdns_xspi, false);
+			cdns_xspi->set_interrupts_handler(cdns_xspi, false);
 			return -EIO;
 		}
 		cdns_xspi->sdma_handler(cdns_xspi);
 	}
 
 	wait_for_completion(&cdns_xspi->cmd_complete);
-	cdns_xspi_set_interrupts(cdns_xspi, false);
+	cdns_xspi->set_interrupts_handler(cdns_xspi, false);
 
 	cmd_status = cdns_xspi_check_command_status(cdns_xspi);
 	if (cmd_status)
@@ -786,9 +804,11 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
 		host->mem_ops = &marvell_xspi_mem_ops;
 		cdns_xspi->sdma_handler = &marvell_xspi_sdma_handle;
+		cdns_xspi->set_interrupts_handler = &marvell_xspi_set_interrupts;
 	} else {
 		host->mem_ops = &cadence_xspi_mem_ops;
 		cdns_xspi->sdma_handler = &cdns_xspi_sdma_handle;
+		cdns_xspi->set_interrupts_handler = &cdns_xspi_set_interrupts;
 	}
 	host->dev.of_node = pdev->dev.of_node;
 	host->bus_num = -1;
-- 
2.43.0

[PATCH v9 6/9] spi: cadence: Add Marvell xfer operation support

[Witold Sadowski]

Marvell Xfer overlay extends xSPI capabilities to support non-memory SPI
operations. The Marvell overlay, combined with a generic command, allows
for full-duplex SPI transactions. It also enables transactions with
undetermined lengths using the cs_hold parameter and the ability to
extend CS signal assertion, even if the xSPI block requests CS signal
de-assertion.
Marvell overlay is using part of xSPI for writing data into device, and
additional hardware block to read data from the device. To do that xSPI
will trigger 1 byte generic command followed by data sequence. In same
time overlay block will monitor MISO pin to read data from the device.
Due to that SDMA data start will be shifted by 1 byte.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 271 +++++++++++++++++++++++++++++++++
 1 file changed, 271 insertions(+)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index eb9293e8d708..fd40f00c3ab0 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -19,6 +19,7 @@
 #include <linux/bitfield.h>
 #include <linux/limits.h>
 #include <linux/log2.h>
+#include <linux/bitrev.h>
 
 #define CDNS_XSPI_MAGIC_NUM_VALUE	0x6522
 #define CDNS_XSPI_MAX_BANKS		8
@@ -193,6 +194,34 @@
 		((op)->data.dir == SPI_MEM_DATA_IN) ? \
 		CDNS_XSPI_STIG_CMD_DIR_READ : CDNS_XSPI_STIG_CMD_DIR_WRITE))
 
+/* Helper macros for GENERIC and GENERIC-DSEQ instruction type */
+#define CMD_REG_LEN (6*4)
+#define INSTRUCTION_TYPE_GENERIC 96
+#define CDNS_XSPI_CMD_FLD_P1_GENERIC_CMD (\
+	FIELD_PREP(CDNS_XSPI_CMD_INSTR_TYPE, INSTRUCTION_TYPE_GENERIC))
+
+#define GENERIC_NUM_OF_BYTES GENMASK(27, 24)
+#define CDNS_XSPI_CMD_FLD_P3_GENERIC_CMD(len) (\
+	FIELD_PREP(GENERIC_NUM_OF_BYTES, len))
+
+#define GENERIC_BANK_NUM GENMASK(14, 12)
+#define GENERIC_GLUE_CMD BIT(28)
+#define CDNS_XSPI_CMD_FLD_P4_GENERIC_CMD(cs, glue) (\
+	FIELD_PREP(GENERIC_BANK_NUM, cs) | FIELD_PREP(GENERIC_GLUE_CMD, glue))
+
+#define CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_1 (\
+	FIELD_PREP(CDNS_XSPI_CMD_INSTR_TYPE, CDNS_XSPI_STIG_INSTR_TYPE_DATA_SEQ))
+
+#define CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_2(nbytes) (\
+	FIELD_PREP(CDNS_XSPI_CMD_DSEQ_R2_DCNT_L, nbytes & 0xffff))
+
+#define CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_3(nbytes) ( \
+	FIELD_PREP(CDNS_XSPI_CMD_DSEQ_R3_DCNT_H, (nbytes >> 16) & 0xffff))
+
+#define CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_4(dir, chipsel) ( \
+	FIELD_PREP(CDNS_XSPI_CMD_DSEQ_R4_BANK, chipsel) | \
+	FIELD_PREP(CDNS_XSPI_CMD_DSEQ_R4_DIR, dir))
+
 /* Marvell PHY default values */
 #define MARVELL_REGS_DLL_PHY_CTRL		0x00000707
 #define MARVELL_CTB_RFILE_PHY_CTRL		0x00004000
@@ -226,6 +255,37 @@
 #define MRVL_XSPI_CLOCK_DIVIDED(div) ((MRVL_XSPI_CLOCK_IO_HZ) / (div))
 #define MRVL_DEFAULT_CLK	     25000000
 
+/* Marvell overlay registers - xfer */
+#define MRVL_XFER_FUNC_CTRL		 0x210
+#define MRVL_XFER_FUNC_CTRL_READ_DATA(i) (0x000 + 8 * (i))
+#define MRVL_XFER_SOFT_RESET		 BIT(11)
+#define MRVL_XFER_CS_N_HOLD		 GENMASK(9, 6)
+#define MRVL_XFER_RECEIVE_ENABLE	 BIT(4)
+#define MRVL_XFER_FUNC_ENABLE		 BIT(3)
+#define MRVL_XFER_CLK_CAPTURE_POL	 BIT(2)
+#define MRVL_XFER_CLK_DRIVE_POL		 BIT(1)
+#define MRVL_XFER_FUNC_START		 BIT(0)
+#define MRVL_XFER_QWORD_COUNT		 32
+#define MRVL_XFER_QWORD_BYTECOUNT	 8
+
+#define MRVL_XSPI_POLL_TIMEOUT_US	1000
+#define MRVL_XSPI_POLL_DELAY_US		10
+
+/* Macros for calculating data bits in generic command
+ * Up to 10 bytes can be fit into cmd_registers
+ * least significant is placed in cmd_reg[1]
+ * Other bits are inserted after it in cmd_reg[1,2,3] register
+ */
+#define GENERIC_CMD_DATA_REG_3_COUNT(len)	(len >= 10 ? 2 : len - 8)
+#define GENERIC_CMD_DATA_REG_2_COUNT(len)	(len >= 7 ? 3 : len - 4)
+#define GENERIC_CMD_DATA_REG_1_COUNT(len)	(len >= 3 ? 2 : len - 1)
+#define GENERIC_CMD_DATA_3_OFFSET(position)	(8*(position))
+#define GENERIC_CMD_DATA_2_OFFSET(position)	(8*(position))
+#define GENERIC_CMD_DATA_1_OFFSET(position)	(8 + 8*(position))
+#define GENERIC_CMD_DATA_INSERT(data, pos)	((data) << (pos))
+#define GENERIC_CMD_REG_3_NEEDED(len)		(len > 7)
+#define GENERIC_CMD_REG_2_NEEDED(len)		(len > 3)
+
 enum cdns_xspi_stig_instr_type {
 	CDNS_XSPI_STIG_INSTR_TYPE_0,
 	CDNS_XSPI_STIG_INSTR_TYPE_1,
@@ -294,6 +354,7 @@ struct cdns_xspi_dev {
 	void __iomem *iobase;
 	void __iomem *auxbase;
 	void __iomem *sdmabase;
+	void __iomem *xferbase;
 
 	int irq;
 	int cur_cs;
@@ -312,6 +373,9 @@ struct cdns_xspi_dev {
 	const struct cdns_xspi_driver_data *driver_data;
 	void (*sdma_handler)(struct cdns_xspi_dev *cdns_xspi);
 	void (*set_interrupts_handler)(struct cdns_xspi_dev *cdns_xspi, bool enabled);
+
+	bool xfer_in_progress;
+	int current_xfer_qword;
 };
 
 static void cdns_xspi_reset_dll(struct cdns_xspi_dev *cdns_xspi)
@@ -780,6 +844,203 @@ static void cdns_xspi_print_phy_config(struct cdns_xspi_dev *cdns_xspi)
 		 readl(cdns_xspi->auxbase + CDNS_XSPI_CCP_PHY_DLL_SLAVE_CTRL));
 }
 
+static int cdns_xspi_prepare_generic(int cs, const void *dout, int len, int glue, u32 *cmd_regs)
+{
+	u8 *data = (u8 *)dout;
+	int i;
+	int data_counter = 0;
+
+	memset(cmd_regs, 0x00, CMD_REG_LEN);
+
+	if (GENERIC_CMD_REG_3_NEEDED(len)) {
+		for (i = GENERIC_CMD_DATA_REG_3_COUNT(len); i >= 0 ; i--)
+			cmd_regs[3] |= GENERIC_CMD_DATA_INSERT(data[data_counter++],
+							       GENERIC_CMD_DATA_3_OFFSET(i));
+	}
+	if (GENERIC_CMD_REG_2_NEEDED(len)) {
+		for (i = GENERIC_CMD_DATA_REG_2_COUNT(len); i >= 0; i--)
+			cmd_regs[2] |= GENERIC_CMD_DATA_INSERT(data[data_counter++],
+							       GENERIC_CMD_DATA_2_OFFSET(i));
+	}
+	for (i = GENERIC_CMD_DATA_REG_1_COUNT(len); i >= 0 ; i--)
+		cmd_regs[1] |= GENERIC_CMD_DATA_INSERT(data[data_counter++],
+						       GENERIC_CMD_DATA_1_OFFSET(i));
+
+	cmd_regs[1] |= CDNS_XSPI_CMD_FLD_P1_GENERIC_CMD;
+	cmd_regs[3] |= CDNS_XSPI_CMD_FLD_P3_GENERIC_CMD(len);
+	cmd_regs[4] |= CDNS_XSPI_CMD_FLD_P4_GENERIC_CMD(cs, glue);
+
+	return 0;
+}
+
+static void marvell_xspi_read_single_qword(struct cdns_xspi_dev *cdns_xspi, u8 **buffer)
+{
+	u64 d = readq(cdns_xspi->xferbase +
+		      MRVL_XFER_FUNC_CTRL_READ_DATA(cdns_xspi->current_xfer_qword));
+	u8 *ptr = (u8 *)&d;
+	int k;
+
+	for (k = 0; k < 8; k++) {
+		u8 val = bitrev8((ptr[k]));
+		**buffer = val;
+		*buffer = *buffer + 1;
+	}
+
+	cdns_xspi->current_xfer_qword++;
+	cdns_xspi->current_xfer_qword %= MRVL_XFER_QWORD_COUNT;
+}
+
+static void cdns_xspi_finish_read(struct cdns_xspi_dev *cdns_xspi, u8 **buffer, u32 data_count)
+{
+	u64 d = readq(cdns_xspi->xferbase +
+		      MRVL_XFER_FUNC_CTRL_READ_DATA(cdns_xspi->current_xfer_qword));
+	u8 *ptr = (u8 *)&d;
+	int k;
+
+	for (k = 0; k < data_count % MRVL_XFER_QWORD_BYTECOUNT; k++) {
+		u8 val = bitrev8((ptr[k]));
+		**buffer = val;
+		*buffer = *buffer + 1;
+	}
+
+	cdns_xspi->current_xfer_qword++;
+	cdns_xspi->current_xfer_qword %= MRVL_XFER_QWORD_COUNT;
+}
+
+static int cdns_xspi_prepare_transfer(int cs, int dir, int len, u32 *cmd_regs)
+{
+	memset(cmd_regs, 0x00, CMD_REG_LEN);
+
+	cmd_regs[1] |= CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_1;
+	cmd_regs[2] |= CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_2(len);
+	cmd_regs[4] |= CDNS_XSPI_CMD_FLD_GENERIC_DSEQ_CMD_4(dir, cs);
+
+	return 0;
+}
+
+static bool cdns_xspi_is_stig_ready(struct cdns_xspi_dev *cdns_xspi, bool sleep)
+{
+	u32 ctrl_stat;
+
+	return !readl_relaxed_poll_timeout
+		(cdns_xspi->iobase + CDNS_XSPI_CTRL_STATUS_REG,
+		ctrl_stat,
+		((ctrl_stat & BIT(3)) == 0),
+		sleep ? MRVL_XSPI_POLL_DELAY_US : 0,
+		sleep ? MRVL_XSPI_POLL_TIMEOUT_US : 0);
+}
+
+static bool cdns_xspi_is_sdma_ready(struct cdns_xspi_dev *cdns_xspi, bool sleep)
+{
+	u32 ctrl_stat;
+
+	return !readl_relaxed_poll_timeout
+		(cdns_xspi->iobase + CDNS_XSPI_INTR_STATUS_REG,
+		ctrl_stat,
+		(ctrl_stat & CDNS_XSPI_SDMA_TRIGGER),
+		sleep ? MRVL_XSPI_POLL_DELAY_US : 0,
+		sleep ? MRVL_XSPI_POLL_TIMEOUT_US : 0);
+}
+
+static int cdns_xspi_transfer_one_message_b0(struct spi_controller *controller,
+					   struct spi_message *m)
+{
+	struct cdns_xspi_dev *cdns_xspi = spi_controller_get_devdata(controller);
+	struct spi_device *spi = m->spi;
+	struct spi_transfer *t = NULL;
+
+	const int max_len = MRVL_XFER_QWORD_BYTECOUNT * MRVL_XFER_QWORD_COUNT;
+	int current_cycle_count;
+	int cs = spi_get_chipselect(spi, 0);
+	int cs_change = 0;
+
+	/* Enable xfer state machine */
+	if (!cdns_xspi->xfer_in_progress) {
+		u32 xfer_control = readl(cdns_xspi->xferbase + MRVL_XFER_FUNC_CTRL);
+
+		cdns_xspi->current_xfer_qword = 0;
+		cdns_xspi->xfer_in_progress = true;
+		xfer_control |= (MRVL_XFER_RECEIVE_ENABLE |
+				 MRVL_XFER_CLK_CAPTURE_POL |
+				 MRVL_XFER_FUNC_START |
+				 MRVL_XFER_SOFT_RESET |
+				 FIELD_PREP(MRVL_XFER_CS_N_HOLD, (1 << cs)));
+		xfer_control &= ~(MRVL_XFER_FUNC_ENABLE | MRVL_XFER_CLK_DRIVE_POL);
+		writel(xfer_control, cdns_xspi->xferbase + MRVL_XFER_FUNC_CTRL);
+	}
+
+	list_for_each_entry(t, &m->transfers, transfer_list) {
+		u8 *txd = (u8 *) t->tx_buf;
+		u8 *rxd = (u8 *) t->rx_buf;
+		u8 data[10];
+		u32 cmd_regs[6];
+
+		if (!txd)
+			txd = data;
+
+		cdns_xspi->in_buffer = txd + 1;
+		cdns_xspi->out_buffer = txd + 1;
+
+		while (t->len) {
+
+			current_cycle_count = t->len > max_len ? max_len : t->len;
+
+			if (current_cycle_count < 10) {
+				cdns_xspi_prepare_generic(cs, txd, current_cycle_count,
+							  false, cmd_regs);
+				cdns_xspi_trigger_command(cdns_xspi, cmd_regs);
+				if (!cdns_xspi_is_stig_ready(cdns_xspi, true))
+					return -EIO;
+			} else {
+				cdns_xspi_prepare_generic(cs, txd, 1, true, cmd_regs);
+				cdns_xspi_trigger_command(cdns_xspi, cmd_regs);
+				cdns_xspi_prepare_transfer(cs, 1, current_cycle_count - 1,
+							   cmd_regs);
+				cdns_xspi_trigger_command(cdns_xspi, cmd_regs);
+				if (!cdns_xspi_is_sdma_ready(cdns_xspi, true))
+					return -EIO;
+				cdns_xspi->sdma_handler(cdns_xspi);
+				if (!cdns_xspi_is_stig_ready(cdns_xspi, true))
+					return -EIO;
+
+				cdns_xspi->in_buffer += current_cycle_count;
+				cdns_xspi->out_buffer += current_cycle_count;
+			}
+
+			if (rxd) {
+				int j;
+
+				for (j = 0; j < current_cycle_count / 8; j++)
+					marvell_xspi_read_single_qword(cdns_xspi, &rxd);
+				cdns_xspi_finish_read(cdns_xspi, &rxd, current_cycle_count);
+			} else {
+				cdns_xspi->current_xfer_qword += current_cycle_count /
+								 MRVL_XFER_QWORD_BYTECOUNT;
+				if (current_cycle_count % MRVL_XFER_QWORD_BYTECOUNT)
+					cdns_xspi->current_xfer_qword++;
+
+				cdns_xspi->current_xfer_qword %= MRVL_XFER_QWORD_COUNT;
+			}
+			cs_change = t->cs_change;
+			t->len -= current_cycle_count;
+		}
+	}
+
+	if (!cs_change) {
+		u32 xfer_control = readl(cdns_xspi->xferbase + MRVL_XFER_FUNC_CTRL);
+
+		xfer_control &= ~(MRVL_XFER_RECEIVE_ENABLE |
+				  MRVL_XFER_SOFT_RESET);
+		writel(xfer_control, cdns_xspi->xferbase + MRVL_XFER_FUNC_CTRL);
+		cdns_xspi->xfer_in_progress = false;
+	}
+
+	m->status = 0;
+	spi_finalize_current_message(controller);
+
+	return 0;
+}
+
 static int cdns_xspi_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
@@ -803,6 +1064,7 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 
 	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
 		host->mem_ops = &marvell_xspi_mem_ops;
+		host->transfer_one_message = cdns_xspi_transfer_one_message_b0;
 		cdns_xspi->sdma_handler = &marvell_xspi_sdma_handle;
 		cdns_xspi->set_interrupts_handler = &marvell_xspi_set_interrupts;
 	} else {
@@ -845,6 +1107,15 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 		return PTR_ERR(cdns_xspi->auxbase);
 	}
 
+	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
+		cdns_xspi->xferbase = devm_platform_ioremap_resource_byname(pdev, "xfer");
+		if (IS_ERR(cdns_xspi->xferbase)) {
+			dev_info(dev, "XFER register base not found, set it\n");
+			// For compatibility with older firmware
+			cdns_xspi->xferbase = cdns_xspi->iobase + 0x8000;
+		}
+	}
+
 	cdns_xspi->irq = platform_get_irq(pdev, 0);
 	if (cdns_xspi->irq < 0)
 		return -ENXIO;
-- 
2.43.0

Re: [PATCH v9 6/9] spi: cadence: Add Marvell xfer operation support

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 844 bytes --]

On Wed, Jun 19, 2024 at 07:17:12AM -0700, Witold Sadowski wrote:

> +		while (t->len) {
> +
> +			current_cycle_count = t->len > max_len ? max_len : t->len;

cycle_count seems like a weird name?

In general please try to avoid overusing the ternery operator, it
doesn't help legibility.  Just write normal conditional statements, or
in this case use max().

> +				cdns_xspi->current_xfer_qword += current_cycle_count /
> +								 MRVL_XFER_QWORD_BYTECOUNT;
> +				if (current_cycle_count % MRVL_XFER_QWORD_BYTECOUNT)
> +					cdns_xspi->current_xfer_qword++;
> +
> +				cdns_xspi->current_xfer_qword %= MRVL_XFER_QWORD_COUNT;
> +			}
> +			cs_change = t->cs_change;
> +			t->len -= current_cycle_count;
> +		}
> +	}

This loop doesn't implement any of the delay stuff, ideally it either
would or would reject transfers that request transfers.

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

RE: [EXTERNAL] Re: [PATCH v9 6/9] spi: cadence: Add Marvell xfer operation support

[Witold Sadowski]

Hi Mark

> 
> > +		while (t->len) {
> > +
> > +			current_cycle_count = t->len > max_len ? max_len : t-
> >len;
> 
> cycle_count seems like a weird name?
> 
> In general please try to avoid overusing the ternery operator, it doesn't
> help legibility.  Just write normal conditional statements, or in this
> case use max().

Will be addressed in next patchset.

> 
> > +				cdns_xspi->current_xfer_qword +=
> current_cycle_count /
> > +
> MRVL_XFER_QWORD_BYTECOUNT;
> > +				if (current_cycle_count %
> MRVL_XFER_QWORD_BYTECOUNT)
> > +					cdns_xspi->current_xfer_qword++;
> > +
> > +				cdns_xspi->current_xfer_qword %=
> MRVL_XFER_QWORD_COUNT;
> > +			}
> > +			cs_change = t->cs_change;
> > +			t->len -= current_cycle_count;
> > +		}
> > +	}
> 
> This loop doesn't implement any of the delay stuff, ideally it either
> would or would reject transfers that request transfers.

I'm not sure about what kind of the delay You are referring to. Can You explain?

Regards
Witek

Re: [EXTERNAL] Re: [PATCH v9 6/9] spi: cadence: Add Marvell xfer operation support

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 320 bytes --]

On Fri, Jun 28, 2024 at 01:48:00PM +0000, Witold Sadowski wrote:

> > This loop doesn't implement any of the delay stuff, ideally it either
> > would or would reject transfers that request transfers.

> I'm not sure about what kind of the delay You are referring to. Can You explain?

The _delay fields in spi_transfer.

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

[PATCH v9 7/9] spi: cadence: Change resource mapping

[Witold Sadowski]

If mapping resource by name will fail try to map resource by number.
Such situation can occur in ACPI case.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
Signed-off-by: Piyush Malgujar <pmalgujar@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 31 ++++++++++++++++++++++---------
 1 file changed, 22 insertions(+), 9 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index fd40f00c3ab0..9cf3624595ea 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -1091,28 +1091,41 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 
 	cdns_xspi->iobase = devm_platform_ioremap_resource_byname(pdev, "io");
 	if (IS_ERR(cdns_xspi->iobase)) {
-		dev_err(dev, "Failed to remap controller base address\n");
-		return PTR_ERR(cdns_xspi->iobase);
+		cdns_xspi->iobase = devm_platform_ioremap_resource(pdev, 0);
+		if (IS_ERR(cdns_xspi->iobase)) {
+			dev_err(dev, "Failed to remap controller base address\n");
+			return PTR_ERR(cdns_xspi->iobase);
+		}
 	}
 
 	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "sdma");
 	cdns_xspi->sdmabase = devm_ioremap_resource(dev, res);
-	if (IS_ERR(cdns_xspi->sdmabase))
-		return PTR_ERR(cdns_xspi->sdmabase);
+	if (IS_ERR(cdns_xspi->sdmabase)) {
+		res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+		cdns_xspi->sdmabase = devm_ioremap_resource(dev, res);
+		if (IS_ERR(cdns_xspi->sdmabase))
+			return PTR_ERR(cdns_xspi->sdmabase);
+	}
 	cdns_xspi->sdmasize = resource_size(res);
 
 	cdns_xspi->auxbase = devm_platform_ioremap_resource_byname(pdev, "aux");
 	if (IS_ERR(cdns_xspi->auxbase)) {
-		dev_err(dev, "Failed to remap AUX address\n");
-		return PTR_ERR(cdns_xspi->auxbase);
+		cdns_xspi->auxbase = devm_platform_ioremap_resource(pdev, 2);
+		if (IS_ERR(cdns_xspi->auxbase)) {
+			dev_err(dev, "Failed to remap AUX address\n");
+			return PTR_ERR(cdns_xspi->auxbase);
+		}
 	}
 
 	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
 		cdns_xspi->xferbase = devm_platform_ioremap_resource_byname(pdev, "xfer");
 		if (IS_ERR(cdns_xspi->xferbase)) {
-			dev_info(dev, "XFER register base not found, set it\n");
-			// For compatibility with older firmware
-			cdns_xspi->xferbase = cdns_xspi->iobase + 0x8000;
+			cdns_xspi->xferbase = devm_platform_ioremap_resource(pdev, 3);
+			if (IS_ERR(cdns_xspi->xferbase)) {
+				dev_info(dev, "XFER register base not found, set it\n");
+				// For compatibility with older firmware
+				cdns_xspi->xferbase = cdns_xspi->iobase + 0x8000;
+			}
 		}
 	}
 
-- 
2.43.0

[PATCH v9 8/9] spi: cadence: Change cs property reading.

[Witold Sadowski]

In current implementation cs property can be read only from
device-tree(for_each_available_child_of_node_scoped). Change it to fwnode
based read to allow property reading in ACPI case too.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
Signed-off-by: Piyush Malgujar <pmalgujar@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index 9cf3624595ea..4edac7c8d73c 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -811,15 +811,20 @@ static irqreturn_t cdns_xspi_irq_handler(int this_irq, void *dev)
 
 static int cdns_xspi_of_get_plat_data(struct platform_device *pdev)
 {
-	struct device_node *node_prop = pdev->dev.of_node;
+	struct fwnode_handle *fwnode_child;
 	unsigned int cs;
 
-	for_each_available_child_of_node_scoped(node_prop, node_child) {
-		if (of_property_read_u32(node_child, "reg", &cs)) {
+	device_for_each_child_node(&pdev->dev, fwnode_child) {
+		if (!fwnode_device_is_available(fwnode_child))
+			continue;
+
+		if (fwnode_property_read_u32(fwnode_child, "reg", &cs)) {
 			dev_err(&pdev->dev, "Couldn't get memory chip select\n");
+			fwnode_handle_put(fwnode_child);
 			return -ENXIO;
 		} else if (cs >= CDNS_XSPI_MAX_BANKS) {
 			dev_err(&pdev->dev, "reg (cs) parameter value too large\n");
+			fwnode_handle_put(fwnode_child);
 			return -ENXIO;
 		}
 	}
-- 
2.43.0

[PATCH v9 9/9] spi: cadence: Try to read spi-tx/rx-bus width property using ACPI

[Witold Sadowski]

Try to read bus width property using acpi_dev_get_property function, do
not rely on spi_mem_default_supports_op function only.
If of_device_get_match_data() will fail, retry with
acpi_device_get_match_data() to handle ACPI properly.

Signed-off-by: Witold Sadowski <wsadowski@marvell.com>
Signed-off-by: Piyush Malgujar <pmalgujar@marvell.com>
---
 drivers/spi/spi-cadence-xspi.c | 74 +++++++++++++++++++++++++++++++++-
 1 file changed, 72 insertions(+), 2 deletions(-)

diff --git a/drivers/spi/spi-cadence-xspi.c b/drivers/spi/spi-cadence-xspi.c
index 4edac7c8d73c..f2d6e8a85e19 100644
--- a/drivers/spi/spi-cadence-xspi.c
+++ b/drivers/spi/spi-cadence-xspi.c
@@ -2,6 +2,7 @@
 // Cadence XSPI flash controller driver
 // Copyright (C) 2020-21 Cadence
 
+#include <linux/acpi.h>
 #include <linux/completion.h>
 #include <linux/delay.h>
 #include <linux/err.h>
@@ -747,6 +748,66 @@ static int marvell_xspi_mem_op_execute(struct spi_mem *mem,
 	return ret;
 }
 
+#ifdef CONFIG_ACPI
+static bool cdns_xspi_supports_op(struct spi_mem *mem,
+				  const struct spi_mem_op *op)
+{
+	struct spi_device *spi = mem->spi;
+	const union acpi_object *obj;
+	struct acpi_device *adev;
+
+	adev = ACPI_COMPANION(&spi->dev);
+
+	if (!acpi_dev_get_property(adev, "spi-tx-bus-width", ACPI_TYPE_INTEGER,
+				   &obj)) {
+		switch (obj->integer.value) {
+		case 1:
+			break;
+		case 2:
+			spi->mode |= SPI_TX_DUAL;
+			break;
+		case 4:
+			spi->mode |= SPI_TX_QUAD;
+			break;
+		case 8:
+			spi->mode |= SPI_TX_OCTAL;
+			break;
+		default:
+			dev_warn(&spi->dev,
+				 "spi-tx-bus-width %lld not supported\n",
+				 obj->integer.value);
+			break;
+		}
+	}
+
+	if (!acpi_dev_get_property(adev, "spi-rx-bus-width", ACPI_TYPE_INTEGER,
+				   &obj)) {
+		switch (obj->integer.value) {
+		case 1:
+			break;
+		case 2:
+			spi->mode |= SPI_RX_DUAL;
+			break;
+		case 4:
+			spi->mode |= SPI_RX_QUAD;
+			break;
+		case 8:
+			spi->mode |= SPI_RX_OCTAL;
+			break;
+		default:
+			dev_warn(&spi->dev,
+				 "spi-rx-bus-width %lld not supported\n",
+				 obj->integer.value);
+			break;
+		}
+	}
+
+	if (!spi_mem_default_supports_op(mem, op))
+		return false;
+
+	return true;
+}
+#endif
 
 static int cdns_xspi_adjust_mem_op_size(struct spi_mem *mem, struct spi_mem_op *op)
 {
@@ -759,11 +820,17 @@ static int cdns_xspi_adjust_mem_op_size(struct spi_mem *mem, struct spi_mem_op *
 }
 
 static const struct spi_controller_mem_ops cadence_xspi_mem_ops = {
+#ifdef CONFIG_ACPI
+	.supports_op = cdns_xspi_supports_op,
+#endif
 	.exec_op = cdns_xspi_mem_op_execute,
 	.adjust_op_size = cdns_xspi_adjust_mem_op_size,
 };
 
 static const struct spi_controller_mem_ops marvell_xspi_mem_ops = {
+#ifdef CONFIG_ACPI
+	.supports_op = cdns_xspi_supports_op,
+#endif
 	.exec_op = marvell_xspi_mem_op_execute,
 	.adjust_op_size = cdns_xspi_adjust_mem_op_size,
 };
@@ -1064,8 +1131,11 @@ static int cdns_xspi_probe(struct platform_device *pdev)
 
 	cdns_xspi = spi_controller_get_devdata(host);
 	cdns_xspi->driver_data = of_device_get_match_data(dev);
-	if (!cdns_xspi->driver_data)
-		return -ENODEV;
+	if (!cdns_xspi->driver_data) {
+		cdns_xspi->driver_data = acpi_device_get_match_data(dev);
+		if (!cdns_xspi->driver_data)
+			return -ENODEV;
+	}
 
 	if (cdns_xspi->driver_data->mrvl_hw_overlay) {
 		host->mem_ops = &marvell_xspi_mem_ops;
-- 
2.43.0