Skip to contentSkip to conversation

ARM i.MX serial: fix tx buffer overflows

PATCH Sascha Hauer 4 participants 5 messages Jan 5, 2007 – Jan 29, 2007 Arm i.MX and Layerscape

lore
Share & tools
lore

Sascha Hauer

author maintainer
#0 lore raw

This patch fixes occasional tx buffer overflows in the i.MX serial driver which came from the fact that space in the buffer was checked after sending the first byte. Also, fifosize is 32 bytes, not 8.

---
 drivers/serial/imx.c |   11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)
drivers/serial/imx.c +5 -6
Index: linux-2.6.20-rc3/drivers/serial/imx.c
===================================================================
@@ -154,7 +154,7 @@ static inline void imx_transmit_buffer(s
 {
 	struct circ_buf *xmit = &sport->port.info->xmit;
 
-	do {
+	while (!(UTS((u32)sport->port.membase) & UTS_TXFULL)) {
 		/* send xmit->buf[xmit->tail]
 		 * out the port here */
 		URTX0((u32)sport->port.membase) = xmit->buf[xmit->tail];
@@ -163,7 +163,7 @@ static inline void imx_transmit_buffer(s
 		sport->port.icount.tx++;
 		if (uart_circ_empty(xmit))
 			break;
-	} while (!(UTS((u32)sport->port.membase) & UTS_TXFULL));
+	}
 
 	if (uart_circ_empty(xmit))
 		imx_stop_tx(&sport->port);
@@ -178,8 +178,7 @@ static void imx_start_tx(struct uart_por
 
 	UCR1((u32)sport->port.membase) |= UCR1_TXMPTYEN;
 
-	if(UTS((u32)sport->port.membase) & UTS_TXEMPTY)
-		imx_transmit_buffer(sport);
+	imx_transmit_buffer(sport);
 }
 
 static irqreturn_t imx_rtsint(int irq, void *dev_id)
@@ -678,7 +677,7 @@ static struct imx_port imx_ports[] = {
 		.mapbase	= IMX_UART1_BASE, /* FIXME */
 		.irq		= UART1_MINT_RX,
 		.uartclk	= 16000000,
-		.fifosize	= 8,
+		.fifosize	= 32,
 		.flags		= UPF_BOOT_AUTOCONF,
 		.ops		= &imx_pops,
 		.line		= 0,
@@ -694,7 +693,7 @@ static struct imx_port imx_ports[] = {
 		.mapbase	= IMX_UART2_BASE, /* FIXME */
 		.irq		= UART2_MINT_RX,
 		.uartclk	= 16000000,
-		.fifosize	= 8,
+		.fifosize	= 32,
 		.flags		= UPF_BOOT_AUTOCONF,
 		.ops		= &imx_pops,
 		.line		= 1,
Acked-by
#1 lore raw
Quoted 5 lines · Sascha Hauer

This patch fixes occasional tx buffer overflows in the i.MX serial driver which came from the fact that space in the buffer was checked after sending the first byte. Also, fifosize is 32 bytes, not 8.

Hello Sascha,

I have tested your change on 2.6.19-rt14 kernel which I have on hand there. It is only very short test, but I have not noticed any problems.

In the fact, I think, that it is possible, that I have noticed some mentioned problem during
my CPU-Freq testing on RT kernels before.
I have noticed some problems sometimes with BusyBox shell history editing during frequency throttling. May it be, that RT+changed frequency invoked the problem.

I have some other small fix for i.MX uart there. Can you add it into patch, or should I send it as separate one-liner?

If RTS interrupt is caused by RTS senzing logic inside i.MX UART module the IRQ type cannot be set.

It applies only for interrupts going through GPIO layer. The problem has been noticed by Konstantin Kletschke some time ago.

No IRQF_TRIGGER set_type function for IRQ 26 (MPU)

I would not change type to fixed 0, because it could be possible to use different GPIO MX1 pin for RTS in the theory. On the other hand it is only for documentation purposes now, because RTS read code would have to be adjusted in such case.

Health and success wishes
in year 2007 from

Pavel Pisa

---

 drivers/serial/imx.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/serial/imx.c +2 -1
Index: linux-2.6.19-rt/drivers/serial/imx.c
===================================================================
@@ -403,7 +403,8 @@ static int imx_startup(struct uart_port 
 	if (retval) goto error_out2;
 
 	retval = request_irq(sport->rtsirq, imx_rtsint,
-			     IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
+			     (sport->rtsirq < IMX_IRQS) ? 0 :
+			       IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
 			     DRIVER_NAME, sport);
 	if (retval) goto error_out3;
 
#2 lore raw

Am 2007-01-05 19:01 +0100 schrieb Pavel Pisa:

Quoted 5 lines

It applies only for interrupts going through GPIO layer. The problem has been noticed by Konstantin Kletschke some time ago.

No IRQF_TRIGGER set_type function for IRQ 26 (MPU)

Yes. I reported this also.

Quoted 7 lines
 drivers/serial/imx.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Index: linux-2.6.19-rt/drivers/serial/imx.c
===================================================================
+++ linux-2.6.19-rt/drivers/serial/imx.c

drivers/serial/imx.c static int imx_startup(struct uart_port · line 409

Quoted 8 lines
drivers/serial/imx.c +2 -1
@@ -403,7 +403,8 @@ static int imx_startup(struct uart_port 
 	if (retval) goto error_out2;
 
 	retval = request_irq(sport->rtsirq, imx_rtsint,
-			     IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
+			     (sport->rtsirq < IMX_IRQS) ? 0 :
+			       IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
 			     DRIVER_NAME, sport);
 	if (retval) goto error_out3;

Okay, this indeed fixes my

No IRQF_TRIGGER set_type function for IRQ 26 (MPU)

here!

Regards, Konsti

Signature · 2 lines
GPG KeyID EF62FCEF
Fingerprint: 13C9 B16B 9844 EC15 CC2E  A080 1E69 3FDA EF62 FCEF
#3 lore raw
Quoted 29 lines · Konstantin Kletschke

Am 2007-01-05 19:01 +0100 schrieb Pavel Pisa:

Quoted 5 lines

It applies only for interrupts going through GPIO layer. The problem has been noticed by Konstantin Kletschke some time ago.

No IRQF_TRIGGER set_type function for IRQ 26 (MPU)

Yes. I reported this also.

Quoted 7 lines
 drivers/serial/imx.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Index: linux-2.6.19-rt/drivers/serial/imx.c
===================================================================
+++ linux-2.6.19-rt/drivers/serial/imx.c

drivers/serial/imx.c static int imx_startup(struct uart_port · line 409

Quoted 8 lines
drivers/serial/imx.c +2 -1
@@ -403,7 +403,8 @@ static int imx_startup(struct uart_port 
 	if (retval) goto error_out2;
 
 	retval = request_irq(sport->rtsirq, imx_rtsint,
-			     IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
+			     (sport->rtsirq < IMX_IRQS) ? 0 :
+			       IRQF_TRIGGER_FALLING | IRQF_TRIGGER_RISING,
 			     DRIVER_NAME, sport);
 	if (retval) goto error_out3;

Okay, this indeed fixes my

Quoted 1 line

No IRQF_TRIGGER set_type function for IRQ 26 (MPU)

Is it really worth adding additional code to shut up this (imho) silly warning? It's just adding needless complexity to drivers.

What happens if a driver is used on multiple platforms, some of which support trigger setting and others which don't? Are we going to end up with a large #ifdef in every driver?

Signature · 3 lines
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:
#4 lore raw

Am 2007-01-29 15:37 +0000 schrieb Russell King:

Quoted 2 lines

Is it really worth adding additional code to shut up this (imho) silly warning? It's just adding needless complexity to drivers.

As I pointed out in
http://lists.arm.linux.org.uk/pipermail/linux-arm-kernel/2006-November/037192.html
the console gets flooded with this "warning" to unusable state sometimes.

Quoted 3 lines

What happens if a driver is used on multiple platforms, some of which support trigger setting and others which don't? Are we going to end up with a large #ifdef in every driver?

I don't know exactly. But in addition to the fact, that this warning floods my console to unusable state I am used to sell software without warnings. If there are warnings my boss some things are broken.

Konsti

Signature · 2 lines
GPG KeyID EF62FCEF
Fingerprint: 13C9 B16B 9844 EC15 CC2E  A080 1E69 3FDA EF62 FCEF