diff mbox series

[v3,3/4] serial: sh-sci: Resume PIO in sci_rx_interrupt() on DMA failure

Message ID 20181210182154.20357-4-geert+renesas@glider.be (mailing list archive)
State New, archived
Headers show
Series serial: sh-sci: Fix fallback to PIO on DMA failure | expand

Commit Message

Geert Uytterhoeven Dec. 10, 2018, 6:21 p.m. UTC
On (H)SCIF, sci_submit_rx() is called in the receive interrupt handler.
Hence if DMA submission fails, the interrupt handler should resume
handling reception using PIO, else no more data is received.

Make sci_submit_rx() return an error indicator, so the receive interrupt
handler can act appropriately.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
v3:
  - Move label handle_pio inside #ifdef to kill defined but not used
    compiler warning when CONFIG_SERIAL_SH_SCI_DMA=n,
  - Move the call sci_submit_rx() in sci_rx_interrupt() up, as it may
    fail, rendering the modification of scr unused,
  - Split in multiple patches.
---
 drivers/tty/serial/sh-sci.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

Comments

Wolfram Sang Dec. 10, 2018, 7 p.m. UTC | #1
Hi Geert,

>  	if (!port_lock_held)
>  		spin_unlock_irqrestore(&port->lock, flags);
> +	return -1;

-Esomething here...

>  		} else {
> +			if (sci_submit_rx(s, false))
> +				goto handle_pio;

and 'sci_submit_rx() < 0' here? I think it makes it more readable that
this is an error case then. 'if sci_submit_rx() do_something' sounds a
bit like the good case if boolean logic was used.

Thanks for picking this up again,

   Wolfram
Geert Uytterhoeven Dec. 11, 2018, 8:16 a.m. UTC | #2
Hi Wolfram,

On Mon, Dec 10, 2018 at 8:00 PM Wolfram Sang <wsa@the-dreams.de> wrote:
> >       if (!port_lock_held)
> >               spin_unlock_irqrestore(&port->lock, flags);
> > +     return -1;
>
> -Esomething here...
>
> >               } else {
> > +                     if (sci_submit_rx(s, false))
> > +                             goto handle_pio;
>
> and 'sci_submit_rx() < 0' here? I think it makes it more readable that
> this is an error case then. 'if sci_submit_rx() do_something' sounds a
> bit like the good case if boolean logic was used.

Thanks, will update for v4, using -EAGAIN.

Gr{oetje,eeting}s,

                        Geert
diff mbox series

Patch

diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
index e353e03ce2602559..d8672336d4ff9b65 100644
--- a/drivers/tty/serial/sh-sci.c
+++ b/drivers/tty/serial/sh-sci.c
@@ -1331,7 +1331,7 @@  static void sci_tx_dma_release(struct sci_port *s)
 	dma_release_channel(chan);
 }
 
-static void sci_submit_rx(struct sci_port *s, bool port_lock_held)
+static int sci_submit_rx(struct sci_port *s, bool port_lock_held)
 {
 	struct dma_chan *chan = s->chan_rx;
 	struct uart_port *port = &s->port;
@@ -1359,7 +1359,7 @@  static void sci_submit_rx(struct sci_port *s, bool port_lock_held)
 	s->active_rx = s->cookie_rx[0];
 
 	dma_async_issue_pending(chan);
-	return;
+	return 0;
 
 fail:
 	/* Switch to PIO */
@@ -1374,6 +1374,7 @@  static void sci_submit_rx(struct sci_port *s, bool port_lock_held)
 	sci_start_rx(port);
 	if (!port_lock_held)
 		spin_unlock_irqrestore(&port->lock, flags);
+	return -1;
 }
 
 static void work_fn_tx(struct work_struct *work)
@@ -1668,8 +1669,10 @@  static irqreturn_t sci_rx_interrupt(int irq, void *ptr)
 			disable_irq_nosync(irq);
 			scr |= SCSCR_RDRQE;
 		} else {
+			if (sci_submit_rx(s, false))
+				goto handle_pio;
+
 			scr &= ~SCSCR_RIE;
-			sci_submit_rx(s, false);
 		}
 		serial_port_out(port, SCSCR, scr);
 		/* Clear current interrupt */
@@ -1681,6 +1684,8 @@  static irqreturn_t sci_rx_interrupt(int irq, void *ptr)
 
 		return IRQ_HANDLED;
 	}
+
+handle_pio:
 #endif
 
 	if (s->rx_trigger > 1 && s->rx_fifo_timeout > 0) {