Skip to contentSkip to conversation

spi/spi_s3c64xx: Fix timeout handling in wait_for_xfer()

PATCH Applied Mark Brown 3 participants 6 messages Sep 7, 2010 – Sep 8, 2010

Subsystems (2)
lore
Share & tools
lore

Mark Brown

author
#0 lore raw

PATCH 1/2 spi/spi_s3c64xx: Fix timeout handling in wait_for_xfer()

In wait_for_xfer() for PIO transfer we are using val as both a counter variable to track the number of spins we've waited for completion and the value we read from the controller, causing us to fail to ever actually notice the timeout. Fix this by using a separate value to hold the register readback.

Also warn when we hit the timeout.

---
 drivers/spi/spi_s3c64xx.c |   11 +++++++----
 1 files changed, 7 insertions(+), 4 deletions(-)
drivers/spi/spi_s3c64xx.c +7 -4
index 6e48ea9..03b28e4 100644
@@ -321,7 +321,7 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
 {
 	struct s3c64xx_spi_info *sci = sdd->cntrlr_info;
 	void __iomem *regs = sdd->regs;
-	unsigned long val;
+	unsigned long val, reg;
 	int ms;
 
 	/* millisecs to xfer 'len' bytes @ 'cur_speed' */
@@ -333,13 +333,16 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
 		val = wait_for_completion_timeout(&sdd->xfer_completion, val);
 	} else {
 		val = msecs_to_loops(ms);
+
 		do {
-			val = readl(regs + S3C64XX_SPI_STATUS);
-		} while (RX_FIFO_LVL(val, sci) < xfer->len && --val);
+			reg = readl(regs + S3C64XX_SPI_STATUS);
+		} while (RX_FIFO_LVL(reg, sci) < xfer->len && --val);
 	}
 
-	if (!val)
+	if (!val) {
+		dev_warn(&sdd->pdev->dev, "Transfer timeout\n");
 		return -EIO;
+	}
 
 	if (dma_mode) {
 		u32 status;
Signature · 1 line
1.7.1
author
#1 lore raw

PATCH 2/2 spi/spi_s3c64xx: Increase dead reckoning time in wait_for_xfer()

For small transfers at high speeds the expected transfer time can easily be well under 1ms, causing the delay in wait_for_xfer() to be only the dead reckoning fudge factor of 5ms currently included. Experiments on some of my systems shows that this is marginal for some transfers so double it to 10ms.

---
 drivers/spi/spi_s3c64xx.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/spi/spi_s3c64xx.c +1 -1
index 03b28e4..4abb441 100644
@@ -326,7 +326,7 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
 
 	/* millisecs to xfer 'len' bytes @ 'cur_speed' */
 	ms = xfer->len * 8 * 1000 / sdd->cur_speed;
-	ms += 5; /* some tolerance */
+	ms += 10; /* some tolerance */
 
 	if (dma_mode) {
 		val = msecs_to_jiffies(ms) + 10;
Signature · 1 line
1.7.1
Acked-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] spi/spi_s3c64xx: Increase dead reckoning time in wait_for_xfer()

Quoted 16 lines · Mark Brown

For small transfers at high speeds the expected transfer time can easily be well under 1ms, causing the delay in wait_for_xfer() to be only the dead reckoning fudge factor of 5ms currently included. Experiments on some of my systems shows that this is marginal for some transfers so double it to 10ms.

---
 drivers/spi/spi_s3c64xx.c |    2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/spi/spi_s3c64xx.c +0 -0
index 03b28e4..4abb441 100644

drivers/spi/spi_s3c64xx.c static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd, · line 332

Quoted 8 lines
drivers/spi/spi_s3c64xx.c +1 -1
@@ -326,7 +326,7 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
 
        /* millisecs to xfer 'len' bytes @ 'cur_speed' */
        ms = xfer->len * 8 * 1000 / sdd->cur_speed;
-       ms += 5; /* some tolerance */
+       ms += 10; /* some tolerance */
 
        if (dma_mode) {
                val = msecs_to_jiffies(ms) + 10;
Applied
#3 lore raw

Re: [PATCH 2/2] spi/spi_s3c64xx: Increase dead reckoning time in wait_for_xfer()

Quoted 24 lines · Jassi Brar
Quoted 22 lines · Mark Brown

For small transfers at high speeds the expected transfer time can easily be well under 1ms, causing the delay in wait_for_xfer() to be only the dead reckoning fudge factor of 5ms currently included. Experiments on some of my systems shows that this is marginal for some transfers so double it to 10ms.

---
 drivers/spi/spi_s3c64xx.c |    2 +-

 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/spi/spi_s3c64xx.c +1 -1
index 03b28e4..4abb441 100644
@@ -326,7 +326,7 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
 
        /* millisecs to xfer 'len' bytes @ 'cur_speed' */
        ms = xfer->len * 8 * 1000 / sdd->cur_speed;
-       ms += 5; /* some tolerance */
+       ms += 10; /* some tolerance */
 
Quoted 4 lines

drivers/spi/spi_s3c64xx.c static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd, · line 332

Quoted 2 lines
drivers/spi/spi_s3c64xx.c +0 -0
        if (dma_mode) {
                val = msecs_to_jiffies(ms) + 10;

Applied, thanks.

g.

#4 lore raw
In reply to Mark Brown
Quoted 33 lines · Mark Brown

In wait_for_xfer() for PIO transfer we are using val as both a counter variable to track the number of spins we've waited for completion and the value we read from the controller, causing us to fail to ever actually notice the timeout. Fix this by using a separate value to hold the register readback.

Also warn when we hit the timeout.

---
 drivers/spi/spi_s3c64xx.c |   11 +++++++----

 1 files changed, 7 insertions(+), 4 deletions(-)
drivers/spi/spi_s3c64xx.c +2 -2
index 6e48ea9..03b28e4 100644
@@ -321,7 +321,7 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
  {
        struct s3c64xx_spi_info *sci = sdd->cntrlr_info;
        void __iomem *regs = sdd->regs;
-       unsigned long val;
+       unsigned long val, reg;
        int ms;
 
        /* millisecs to xfer 'len' bytes @ 'cur_speed' */
@@ -333,13 +333,16 @@ static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd,
                val = wait_for_completion_timeout(&sdd->xfer_completion, val);
        } else {
                val = msecs_to_loops(ms);
+
                do {
-                       val = readl(regs + S3C64XX_SPI_STATUS);

drivers/spi/spi_s3c64xx.c static int wait_for_xfer(struct s3c64xx_spi_driver_data *sdd, · line 345

Quoted 10 lines
drivers/spi/spi_s3c64xx.c +5 -2
-               } while (RX_FIFO_LVL(val, sci) < xfer->len && --val);
+                       reg = readl(regs + S3C64XX_SPI_STATUS);
+               } while (RX_FIFO_LVL(reg, sci) < xfer->len && --val);
        }
 
-       if (!val)
+       if (!val) {
+               dev_warn(&sdd->pdev->dev, "Transfer timeout\n");
                return -EIO;
+       }

I have already submitted a patch a few days ago
https://patchwork.kernel.org/patch/151941/
(It's strange that Grant's id isn't there in the CC list, despite my writing it)

author
#5 lore raw
Quoted 3 lines · Jassi Brar

I have already submitted a patch a few days ago
https://patchwork.kernel.org/patch/151941/
(It's strange that Grant's id isn't there in the CC list, despite my writing it)

Ah, excellent. Now I see patchwork there look to be some other things sitting in patchwork which might be handy for me as well.