On 11/29/2013 06:00 PM, Marek Vasut wrote:
Dear Hector Palacios,
The ISR was handling CTS flag and RX/TX processing.
Occasionally interrupts were coming in the middle of the ISR and
led to blocking transmission (with hw flow) or loss of data.
This patch moves this processing to a tasklet that the ISR schedules.
Signed-off-by: Hector Palacios <hector.palacios@xxxxxxxx>
---
drivers/tty/serial/mxs-auart.c | 61
+++++++++++++++++++++++++++--------------- 1 file changed, 39
insertions(+), 22 deletions(-)
diff --git a/drivers/tty/serial/mxs-auart.c
b/drivers/tty/serial/mxs-auart.c index e5540c89dd48..7021269c70b2 100644
--- a/drivers/tty/serial/mxs-auart.c
+++ b/drivers/tty/serial/mxs-auart.c
@@ -148,6 +148,10 @@ struct mxs_auart_port {
struct clk *clk;
struct device *dev;
+ struct tasklet_struct tasklet;
+ unsigned int irq_status;
+ unsigned int status;
+
/* for DMA */
struct scatterlist tx_sgl;
struct dma_chan *tx_dma_chan;
@@ -680,38 +684,48 @@ static void mxs_auart_settermios(struct uart_port *u,
}
}
+/*
+ * tasklet handling tty stuff outside the interrupt handler.
+ */
+static void mxs_auart_tasklet_func(unsigned long data)
+{
+ struct uart_port *port = (struct uart_port *)data;
+ struct mxs_auart_port *s = to_auart_port(port);
+
+ /* Handle status irq */
+ if (s->irq_status & AUART_INTR_CTSMIS) {
+ uart_handle_cts_change(&s->port, s->status & AUART_STAT_CTS);
+ writel(AUART_INTR_CTSMIS,
+ s->port.membase + AUART_INTR_CLR);
+ }
+ /* Handle receive irq */
+ if (s->irq_status & (AUART_INTR_RTIS | AUART_INTR_RXIS)) {
+ if (!auart_dma_enabled(s))
+ mxs_auart_rx_chars(s);
+ }
+ /* Handle transmit irq */
+ if (s->irq_status & AUART_INTR_TXIS)
+ mxs_auart_tx_chars(s);
+}
+
static irqreturn_t mxs_auart_irq_handle(int irq, void *context)
{
- u32 istat;
struct mxs_auart_port *s = context;
- u32 stat = readl(s->port.membase + AUART_STAT);
- istat = readl(s->port.membase + AUART_INTR);
+ s->status = readl(s->port.membase + AUART_STAT);
+ s->irq_status = readl(s->port.membase + AUART_INTR);
Aren't you up for a nice race condition here if you get interrupt while checking
the s->irq_status bits in the tasklet ?
Sad, but true. :-(
But if I need to wait on the ISR for the tasklet to process then this patch loses its
original meaning. Argh... it works so nicely!
Best regards,
--
Hector Palacios
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
