[PATCH/RFC] serial: sh-sci: Fix issue of frequent rx_full interrupt

[PATCH/RFC] serial: sh-sci: Fix issue of frequent rx_full interrupt

[Yoshihiro Kaneko]

From: Kazuya Mizuguchi <kazuya.mizuguchi.ks@renesas.com>

When the receive FIFO is not read in PIO transfer, rx_full interrupt
occurs frequently. This patch avoids it by discarding the receive FIFO.

Signed-off-by: Kazuya Mizuguchi <kazuya.mizuguchi.ks@renesas.com>
Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
---

This patch is based on the tty-next branch of Greg Kroah-Hartman's tty
tree.

 drivers/tty/serial/sh-sci.c | 22 ++++++++++++++++++++--
 1 file changed, 20 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
index b74a644..4ec3c32 100644
--- a/drivers/tty/serial/sh-sci.c
+++ b/drivers/tty/serial/sh-sci.c
@@ -121,6 +121,8 @@ struct sci_port {
 #endif
 
 	struct notifier_block		freq_transition;
+
+	u64				rxfull_time;
 };
 
 /* Function prototypes */
@@ -690,12 +692,26 @@ static void sci_receive_chars(struct uart_port *port)
 		return;
 
 	while (1) {
+		int rxfill = sci_rxfill(port);
+		u64 now = get_jiffies_64();
 		/* Don't copy more bytes than there is room for in the buffer */
-		count = tty_buffer_request_room(tport, sci_rxfill(port));
+		count = tty_buffer_request_room(tport, rxfill);
 
 		/* If for any reason we can't copy more data, we're done! */
-		if (count = 0)
+		if (count = 0) {
+			int n;
+			/* discarded bytes in RX-FIFO */
+			for (n = 0; n < rxfill; n++)
+				serial_port_in(port, SCxRDR);
+			if (rxfill > 0 &&
+			    time_after64(now, sci_port->rxfull_time + (5*HZ))) {
+				sci_port->rxfull_time = now;
+				dev_warn(port->dev,
+					 "RX-FIFO isn't read, so %d bytes is discarded.\n",
+					 rxfill);
+			}
 			break;
+		}
 
 		if (port->type = PORT_SCI) {
 			char c = serial_port_in(port, SCxRDR);
@@ -2342,6 +2358,8 @@ static int sci_init_single(struct platform_device *dev,
 		dev_dbg(port->dev, "DMA tx %d, rx %d\n",
 			p->dma_slave_tx, p->dma_slave_rx);
 
+	sci_port->rxfull_time = get_jiffies_64();
+
 	return 0;
 }
 
-- 
1.9.1

Re: [PATCH/RFC] serial: sh-sci: Fix issue of frequent rx_full interrupt

[Geert Uytterhoeven]

Hi Kaneko-san, Mizuguchi-san,

On Sun, Jun 14, 2015 at 7:26 PM, Yoshihiro Kaneko <ykaneko0929@gmail.com> wrote:
> From: Kazuya Mizuguchi <kazuya.mizuguchi.ks@renesas.com>
>
> When the receive FIFO is not read in PIO transfer, rx_full interrupt
> occurs frequently. This patch avoids it by discarding the receive FIFO.
>
> Signed-off-by: Kazuya Mizuguchi <kazuya.mizuguchi.ks@renesas.com>
> Signed-off-by: Yoshihiro Kaneko <ykaneko0929@gmail.com>
> ---
>
> This patch is based on the tty-next branch of Greg Kroah-Hartman's tty
> tree.
>
>  drivers/tty/serial/sh-sci.c | 22 ++++++++++++++++++++--
>  1 file changed, 20 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
> index b74a644..4ec3c32 100644
> --- a/drivers/tty/serial/sh-sci.c
> +++ b/drivers/tty/serial/sh-sci.c
> @@ -121,6 +121,8 @@ struct sci_port {
>  #endif
>
>         struct notifier_block           freq_transition;
> +
> +       u64                             rxfull_time;
>  };
>
>  /* Function prototypes */
> @@ -690,12 +692,26 @@ static void sci_receive_chars(struct uart_port *port)
>                 return;
>
>         while (1) {
> +               int rxfill = sci_rxfill(port);
> +               u64 now = get_jiffies_64();
>                 /* Don't copy more bytes than there is room for in the buffer */
> -               count = tty_buffer_request_room(tport, sci_rxfill(port));
> +               count = tty_buffer_request_room(tport, rxfill);
>
>                 /* If for any reason we can't copy more data, we're done! */
> -               if (count = 0)
> +               if (count = 0) {
> +                       int n;
> +                       /* discarded bytes in RX-FIFO */
> +                       for (n = 0; n < rxfill; n++)
> +                               serial_port_in(port, SCxRDR);
> +                       if (rxfill > 0 &&
> +                           time_after64(now, sci_port->rxfull_time + (5*HZ))) {
> +                               sci_port->rxfull_time = now;
> +                               dev_warn(port->dev,
> +                                        "RX-FIFO isn't read, so %d bytes is discarded.\n",
> +                                        rxfill);

If I'm not mistaken, this means we loose data, without the tty user noticing?
As data is discarded from the FIFO, we won't get an overrun condition later,
and TTY_OVERRUN will never be inserted into the flip buffer.

> +                       }
>                         break;
> +               }
>
>                 if (port->type = PORT_SCI) {
>                         char c = serial_port_in(port, SCxRDR);
> @@ -2342,6 +2358,8 @@ static int sci_init_single(struct platform_device *dev,
>                 dev_dbg(port->dev, "DMA tx %d, rx %d\n",
>                         p->dma_slave_tx, p->dma_slave_rx);
>
> +       sci_port->rxfull_time = get_jiffies_64();
> +
>         return 0;
>  }

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds