On 10/10/2017 11:22 PM, Han, Nandor (GE Healthcare) wrote:

quoted

-----Original Message-----
From: linux-serial-owner at vger.kernel.org [mailto:linux-serial-
owner at vger.kernel.org] On Behalf Of Troy Kisky
Sent: 07 October 2017 04:47
To: gregkh at linuxfoundation.org
Cc: fabio.estevam at nxp.com; l.stach at pengutronix.de; u.kleine-
koenig at pengutronix.de; linux-serial at vger.kernel.org; linux-arm-
kernel at lists.infradead.org; Troy Kisky [off-list ref]
Subject: EXT: [PATCH 1/1] tty: serial: imx: disable ageing timer interrupt if
dma in use

Since commit 4dec2f119e86 ("imx-serial: RX DMA startup latency")

the interrupt routine no longer will start rx dma.

I think you have 2 commits here: 
- one that disables ageing timer when DMA is in use
- one that cleans the leftovers after 4dec2f119e86.

I could kill 3 lines, in the 1st patch, and the other 27 in the next.
But the 2 patches would not be independent. You should not apply the
second without the 1st. I think the change is simple enough not to
warrant a series.

quoted

Signed-off-by: Troy Kisky <redacted>
---
 drivers/tty/serial/imx.c | 34 ++--------------------------------
 1 file changed, 2 insertions(+), 32 deletions(-)

diff --git a/drivers/tty/serial/imx.c b/drivers/tty/serial/imx.c index

dfeff3951f93..15b0ecb4cf60 100644

--- a/drivers/tty/serial/imx.c
+++ b/drivers/tty/serial/imx.c

@@ -732,29 +732,6 @@ static void imx_disable_rx_int(struct imx_port

*sport)  }

 static void clear_rx_errors(struct imx_port *sport); -static int
start_rx_dma(struct imx_port *sport);
-/*
- * If the RXFIFO is filled with some data, and then we
- * arise a DMA operation to receive them.
- */
-static void imx_dma_rxint(struct imx_port *sport) -{
-	unsigned long temp;
-	unsigned long flags;
-
-	spin_lock_irqsave(&sport->port.lock, flags);
-
-	temp = readl(sport->port.membase + USR2);
-	if ((temp & USR2_RDR) && !sport->dma_is_rxing) {
-
-		imx_disable_rx_int(sport);
-
-		/* tell the DMA to receive the data. */
-		start_rx_dma(sport);
-	}
-
-	spin_unlock_irqrestore(&sport->port.lock, flags);
-}

 /*
  * We have a modem side uart, so the meanings of RTS and CTS are inverted.

@@ -816,11 +793,8 @@ static irqreturn_t imx_int(int irq, void *dev_id)
 	sts = readl(sport->port.membase + USR1);
 	sts2 = readl(sport->port.membase + USR2);

-	if (sts & (USR1_RRDY | USR1_AGTIM)) {
-		if (sport->dma_is_enabled)
-			imx_dma_rxint(sport);
-		else
-			imx_rxint(irq, dev_id);
+	if (!sport->dma_is_enabled && (sts & (USR1_RRDY |

USR1_AGTIM))) {
+		imx_rxint(irq, dev_id);

I don't think we need the check `!sport->dma_is_enabled ` since RRDY and AGTIM interrupts should be disable when DMA is enabled.

If I AND the status with its corresponding interrupt enable. But that is just as complicated, and
more different from the original.

quoted

 		ret = IRQ_HANDLED;
 	}

@@ -1207,10 +1181,6 @@ static void imx_enable_dma(struct imx_port

*sport)
 	temp |= UCR1_RDMAEN | UCR1_TDMAEN | UCR1_ATDMAEN;
 	writel(temp, sport->port.membase + UCR1);

-	temp = readl(sport->port.membase + UCR2);
-	temp |= UCR2_ATEN;
-	writel(temp, sport->port.membase + UCR2);
-

I think this should be moved to imx_startup. At least RRDY is configured there. 

Currently, non-dma mode does not use the ageing timer. You get an interrupt immediately when
1 character is in the fifo. So, enabling the ageing timer and changing fifo level, should really be
a separate patch. But it would be a nice patch.