Skip to contentSkip to conversation

spi: dw: add check for Rx FIFO overflow

PATCH Superseded Maksim Kiselev 2 participants 4 messages Oct 17, 2023 – Dec 20, 2023 SPI

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 spi: dw: add check for Rx FIFO overflow

If even one byte is lost due to Rx FIFO overflow then we will never exit the read loop. Because the (priv->rx != priv->rx_end) condition will be always true.

Let's check if Rx FIFO overflow occurred and exit the read loop in this case.

---
 drivers/spi/designware_spi.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)
drivers/spi/designware_spi.c +21 -3
index 1c7d0ca310..0f443bff8e 100644
@@ -111,6 +111,15 @@
 #define SR_TX_ERR			BIT(5)
 #define SR_DCOL				BIT(6)
 
+/* Bit fields in ISR, IMR, RISR, 7 bits */
+#define DW_SPI_INT_MASK			GENMASK(5, 0)
+#define DW_SPI_INT_TXEI			BIT(0)
+#define DW_SPI_INT_TXOI			BIT(1)
+#define DW_SPI_INT_RXUI			BIT(2)
+#define DW_SPI_INT_RXOI			BIT(3)
+#define DW_SPI_INT_RXFI			BIT(4)
+#define DW_SPI_INT_MSTI			BIT(5)
+
 #define RX_TIMEOUT			1000		/* timeout in ms */
 
 struct dw_spi_plat {
@@ -588,7 +597,7 @@ static int dw_spi_exec_op(struct spi_slave *slave, const struct spi_mem_op *op)
 	struct dw_spi_priv *priv = dev_get_priv(bus);
 	u8 op_len = op->cmd.nbytes + op->addr.nbytes + op->dummy.nbytes;
 	u8 op_buf[op_len];
-	u32 cr0;
+	u32 cr0, sts;
 
 	if (read)
 		priv->tmode = CTRLR0_TMOD_EPROMREAD;
@@ -632,12 +641,21 @@ static int dw_spi_exec_op(struct spi_slave *slave, const struct spi_mem_op *op)
 	 * them to fail because we are not reading/writing the fifo fast enough.
 	 */
 	if (read) {
-		priv->rx = op->data.buf.in;
+		void *prev_rx = priv->rx = op->data.buf.in;
 		priv->rx_end = priv->rx + op->data.nbytes;
 
 		dw_write(priv, DW_SPI_SER, 1 << spi_chip_select(slave->dev));
-		while (priv->rx != priv->rx_end)
+		while (priv->rx != priv->rx_end) {
 			dw_reader(priv);
+			if (prev_rx == priv->rx) {
+				sts = dw_read(priv, DW_SPI_RISR);
+				if (sts & DW_SPI_INT_RXOI) {
+					dev_err(bus, "FIFO overflow on Rx\n");
+					return -EIO;
+				}
+			}
+			prev_rx = priv->rx;
+		}
 	} else {
 		u32 val;
 
Signature · 1 line
2.40.1
#1 lore raw
Quoted 19 lines · Maksim Kiselev

If even one byte is lost due to Rx FIFO overflow then we will never exit the read loop. Because the (priv->rx != priv->rx_end) condition will be always true.

Let's check if Rx FIFO overflow occurred and exit the read loop in this case.

---
 drivers/spi/designware_spi.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)
drivers/spi/designware_spi.c +0 -0
index 1c7d0ca310..0f443bff8e 100644
@@ -111,6 +111,15 @@
 #define SR_TX_ERR                      BIT(5)

drivers/spi/designware_spi.c line 121

Quoted 10 lines
drivers/spi/designware_spi.c +8 -0
 #define SR_DCOL                                BIT(6)
 
+/* Bit fields in ISR, IMR, RISR, 7 bits */
+#define DW_SPI_INT_MASK                        GENMASK(5, 0)
+#define DW_SPI_INT_TXEI                        BIT(0)
+#define DW_SPI_INT_TXOI                        BIT(1)
+#define DW_SPI_INT_RXUI                        BIT(2)
+#define DW_SPI_INT_RXOI                        BIT(3)
+#define DW_SPI_INT_RXFI                        BIT(4)
+#define DW_SPI_INT_MSTI                        BIT(5)

Why do we need unused macros?

Jagan.

#2 lore raw

Hello Jagan,

пн, 18 дек. 2023 г. в 14:28, Jagan Teki <[email protected]>:

Quoted 23 lines · Maksim Kiselev
Quoted 21 lines

If even one byte is lost due to Rx FIFO overflow then we will never exit the read loop. Because the (priv->rx != priv->rx_end) condition will be always true.

Let's check if Rx FIFO overflow occurred and exit the read loop in this case.

---
 drivers/spi/designware_spi.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)
drivers/spi/designware_spi.c +0 -0
index 1c7d0ca310..0f443bff8e 100644
@@ -111,6 +111,15 @@
 #define SR_TX_ERR                      BIT(5)
 #define SR_DCOL                                BIT(6)
 
Quoted 10 lines

drivers/spi/designware_spi.c line 121

Quoted 8 lines
drivers/spi/designware_spi.c +8 -0
+/* Bit fields in ISR, IMR, RISR, 7 bits */
+#define DW_SPI_INT_MASK                        GENMASK(5, 0)
+#define DW_SPI_INT_TXEI                        BIT(0)
+#define DW_SPI_INT_TXOI                        BIT(1)
+#define DW_SPI_INT_RXUI                        BIT(2)
+#define DW_SPI_INT_RXOI                        BIT(3)
+#define DW_SPI_INT_RXFI                        BIT(4)
+#define DW_SPI_INT_MSTI                        BIT(5)

Why do we need unused macros?

Actually DW_SPI_INT_RXOI is used. As for the other bits in DW_SPI_RISR, I decided to add them all to match a Linux driver.

We already have a lot of unused macro in this driver (ex. definitions for bits in CTRLR0_FRF,
CTRLR0_MODE regs and so on).

So, what's a problem with adding the DW_SPI_RISR bits?

Cheers,
Maksim

Quoted 2 lines

Jagan.

#3 lore raw
Quoted 38 lines · Maxim Kiselev

Hello Jagan,

пн, 18 дек. 2023 г. в 14:28, Jagan Teki <[email protected]>:

Quoted 23 lines · Maksim Kiselev
Quoted 21 lines

If even one byte is lost due to Rx FIFO overflow then we will never exit the read loop. Because the (priv->rx != priv->rx_end) condition will be always true.

Let's check if Rx FIFO overflow occurred and exit the read loop in this case.

---
 drivers/spi/designware_spi.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)
drivers/spi/designware_spi.c +0 -0
index 1c7d0ca310..0f443bff8e 100644
@@ -111,6 +111,15 @@
 #define SR_TX_ERR                      BIT(5)
 #define SR_DCOL                                BIT(6)
 
Quoted 10 lines

drivers/spi/designware_spi.c line 121

Quoted 8 lines
drivers/spi/designware_spi.c +8 -0
+/* Bit fields in ISR, IMR, RISR, 7 bits */
+#define DW_SPI_INT_MASK                        GENMASK(5, 0)
+#define DW_SPI_INT_TXEI                        BIT(0)
+#define DW_SPI_INT_TXOI                        BIT(1)
+#define DW_SPI_INT_RXUI                        BIT(2)
+#define DW_SPI_INT_RXOI                        BIT(3)
+#define DW_SPI_INT_RXFI                        BIT(4)
+#define DW_SPI_INT_MSTI                        BIT(5)

Why do we need unused macros?

Quoted 8 lines

Actually DW_SPI_INT_RXOI is used. As for the other bits in DW_SPI_RISR, I decided to add them all to match a Linux driver.

We already have a lot of unused macro in this driver (ex. definitions for bits in CTRLR0_FRF,
CTRLR0_MODE regs and so on).

So, what's a problem with adding the DW_SPI_RISR bits?

I'm not fond of unused change in the particular patch, better to send the updated one with used bits.

Jagan.