Merge tag 'spi-v5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi

Pull spi updates from Mark Brown:
 "The big theme for this release has been performance, we've had a
  series of unrelated overhauls of a few drivers all with a big
  peformance component.

  Otherwise it's been relatively quiet, highlights include:

   - A big overhaul of the spi-fsl-dspi driver improving the code
     quality, performance and stability from Vladimir Oltean.

   - A big performance enhancement for the bc2835 (Raspberry Pi) driver
     for unidirectional transfers from Lukas Wunner.

   - Improved performance on small transfers for the uniphier driver
     from Keiji Hayashibara.

   - Lots of coccinelle generated cleanups from Yue Haibing.

   - New device support for Freescale ls2080a and Nuvoton NPCM FIU"

* tag 'spi-v5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi: (102 commits)
  spi: mediatek: support large PA
  spi: mediatek: add spi support for mt6765 IC
  dt-bindings: spi: update bindings for MT6765 SoC
  spi: bcm2835: Speed up RX-only DMA transfers by zero-filling TX FIFO
  spi: bcm2835: Speed up TX-only DMA transfers by clearing RX FIFO
  dmaengine: bcm2835: Avoid accessing memory when copying zeroes
  spi: bcm2835: Cache CS register value for ->prepare_message()
  dmaengine: bcm2835: Document struct bcm2835_dmadev
  spi: Guarantee cacheline alignment of driver-private data
  dmaengine: bcm2835: Allow reusable descriptors
  dmaengine: bcm2835: Allow cyclic transactions without interrupt
  spi: bcm2835: Drop dma_pending flag
  spi: bcm2835: Work around DONE bit erratum
  spi-gpio: Use PTR_ERR_OR_ZERO() in spi_gpio_request()
  spi: Use an abbreviated pointer to ctlr->cur_msg in __spi_pump_messages
  spi: npcm-fiu: remove set but not used variable 'retlen'
  spi: fsl-spi: use devm_platform_ioremap_resource() to simplify code
  spi: zynq-qspi: use devm_platform_ioremap_resource() to simplify code
  spi: zynqmp: use devm_platform_ioremap_resource() to simplify code
  spi: xlp: use devm_platform_ioremap_resource() to simplify code
  ...

1 files changed, 69 insertions, 21 deletions

diff --git a/drivers/spi/spi-uniphier.c b/drivers/spi/spi-uniphier.c
index b32c77df5d49..47cde1864630 100644
--- a/drivers/spi/spi-uniphier.c
+++ b/drivers/spi/spi-uniphier.c
@@ -7,6 +7,7 @@
 #include <linux/bitfield.h>
 #include <linux/bitops.h>
 #include <linux/clk.h>
+#include <linux/delay.h>
 #include <linux/interrupt.h>
 #include <linux/io.h>
 #include <linux/module.h>
@@ -16,6 +17,7 @@
 #include <asm/unaligned.h>
 
 #define SSI_TIMEOUT_MS		2000
+#define SSI_POLL_TIMEOUT_US	200
 #define SSI_MAX_CLK_DIVIDER	254
 #define SSI_MIN_CLK_DIVIDER	4
 
@@ -214,6 +216,7 @@ static void uniphier_spi_setup_transfer(struct spi_device *spi,
 	if (!priv->is_save_param || priv->mode != spi->mode) {
 		uniphier_spi_set_mode(spi);
 		priv->mode = spi->mode;
+		priv->is_save_param = false;
 	}
 
 	if (!priv->is_save_param || priv->bits_per_word != t->bits_per_word) {
@@ -226,8 +229,7 @@ static void uniphier_spi_setup_transfer(struct spi_device *spi,
 		priv->speed_hz = t->speed_hz;
 	}
 
-	if (!priv->is_save_param)
-		priv->is_save_param = true;
+	priv->is_save_param = true;
 
 	/* reset FIFOs */
 	val = SSI_FC_TXFFL | SSI_FC_RXFFL;
@@ -290,21 +292,23 @@ static void uniphier_spi_recv(struct uniphier_spi_priv *priv)
 
 static void uniphier_spi_fill_tx_fifo(struct uniphier_spi_priv *priv)
 {
-	unsigned int tx_count;
+	unsigned int fifo_threshold, fill_bytes;
 	u32 val;
 
-	tx_count = DIV_ROUND_UP(priv->tx_bytes,
+	fifo_threshold = DIV_ROUND_UP(priv->rx_bytes,
 				bytes_per_word(priv->bits_per_word));
-	tx_count = min(tx_count, SSI_FIFO_DEPTH);
+	fifo_threshold = min(fifo_threshold, SSI_FIFO_DEPTH);
+
+	fill_bytes = fifo_threshold - (priv->rx_bytes - priv->tx_bytes);
 
 	/* set fifo threshold */
 	val = readl(priv->base + SSI_FC);
 	val &= ~(SSI_FC_TXFTH_MASK | SSI_FC_RXFTH_MASK);
-	val |= FIELD_PREP(SSI_FC_TXFTH_MASK, tx_count);
-	val |= FIELD_PREP(SSI_FC_RXFTH_MASK, tx_count);
+	val |= FIELD_PREP(SSI_FC_TXFTH_MASK, fifo_threshold);
+	val |= FIELD_PREP(SSI_FC_RXFTH_MASK, fifo_threshold);
 	writel(val, priv->base + SSI_FC);
 
-	while (tx_count--)
+	while (fill_bytes--)
 		uniphier_spi_send(priv);
 }
 
@@ -323,20 +327,14 @@ static void uniphier_spi_set_cs(struct spi_device *spi, bool enable)
 	writel(val, priv->base + SSI_FPS);
 }
 
-static int uniphier_spi_transfer_one(struct spi_master *master,
-				     struct spi_device *spi,
-				     struct spi_transfer *t)
+static int uniphier_spi_transfer_one_irq(struct spi_master *master,
+					 struct spi_device *spi,
+					 struct spi_transfer *t)
 {
 	struct uniphier_spi_priv *priv = spi_master_get_devdata(master);
 	struct device *dev = master->dev.parent;
 	unsigned long time_left;
 
-	/* Terminate and return success for 0 byte length transfer */
-	if (!t->len)
-		return 0;
-
-	uniphier_spi_setup_transfer(spi, t);
-
 	reinit_completion(&priv->xfer_done);
 
 	uniphier_spi_fill_tx_fifo(priv);
@@ -356,6 +354,59 @@ static int uniphier_spi_transfer_one(struct spi_master *master,
 	return priv->error;
 }
 
+static int uniphier_spi_transfer_one_poll(struct spi_master *master,
+					  struct spi_device *spi,
+					  struct spi_transfer *t)
+{
+	struct uniphier_spi_priv *priv = spi_master_get_devdata(master);
+	int loop = SSI_POLL_TIMEOUT_US * 10;
+
+	while (priv->tx_bytes) {
+		uniphier_spi_fill_tx_fifo(priv);
+
+		while ((priv->rx_bytes - priv->tx_bytes) > 0) {
+			while (!(readl(priv->base + SSI_SR) & SSI_SR_RNE)
+								&& loop--)
+				ndelay(100);
+
+			if (loop == -1)
+				goto irq_transfer;
+
+			uniphier_spi_recv(priv);
+		}
+	}
+
+	return 0;
+
+irq_transfer:
+	return uniphier_spi_transfer_one_irq(master, spi, t);
+}
+
+static int uniphier_spi_transfer_one(struct spi_master *master,
+				     struct spi_device *spi,
+				     struct spi_transfer *t)
+{
+	struct uniphier_spi_priv *priv = spi_master_get_devdata(master);
+	unsigned long threshold;
+
+	/* Terminate and return success for 0 byte length transfer */
+	if (!t->len)
+		return 0;
+
+	uniphier_spi_setup_transfer(spi, t);
+
+	/*
+	 * If the transfer operation will take longer than
+	 * SSI_POLL_TIMEOUT_US, it should use irq.
+	 */
+	threshold = DIV_ROUND_UP(SSI_POLL_TIMEOUT_US * priv->speed_hz,
+					USEC_PER_SEC * BITS_PER_BYTE);
+	if (t->len > threshold)
+		return uniphier_spi_transfer_one_irq(master, spi, t);
+	else
+		return uniphier_spi_transfer_one_poll(master, spi, t);
+}
+
 static int uniphier_spi_prepare_transfer_hardware(struct spi_master *master)
 {
 	struct uniphier_spi_priv *priv = spi_master_get_devdata(master);
@@ -419,7 +470,6 @@ static int uniphier_spi_probe(struct platform_device *pdev)
 {
 	struct uniphier_spi_priv *priv;
 	struct spi_master *master;
-	struct resource *res;
 	unsigned long clk_rate;
 	int irq;
 	int ret;
@@ -434,8 +484,7 @@ static int uniphier_spi_probe(struct platform_device *pdev)
 	priv->master = master;
 	priv->is_save_param = false;
 
-	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
-	priv->base = devm_ioremap_resource(&pdev->dev, res);
+	priv->base = devm_platform_ioremap_resource(pdev, 0);
 	if (IS_ERR(priv->base)) {
 		ret = PTR_ERR(priv->base);
 		goto out_master_put;
@@ -454,7 +503,6 @@ static int uniphier_spi_probe(struct platform_device *pdev)
 
 	irq = platform_get_irq(pdev, 0);
 	if (irq < 0) {
-		dev_err(&pdev->dev, "failed to get IRQ\n");
 		ret = irq;
 		goto out_disable_clk;
 	}