serial: sh-sci: let timeout timer only run when DMA is scheduled
authorWolfram Sang <wsa+renesas@sang-engineering.com>
Mon, 6 May 2024 11:40:19 +0000 (13:40 +0200)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 10 May 2024 10:48:33 +0000 (11:48 +0100)
The hrtimer for RXDMA timeout was unconditionally restarted in the RXDMA
complete handler ignoring the fact that setting up DMA may fail and PIO
is used instead. Explicitly stop the timer when DMA is completed and
only restart it when setting up DMA was successful. This makes the
intention of the timer much clearer, the driver easier to understand and
simplifies assumptions about the timer. The latter avoids race
conditions if these assumptions were not met or confused.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Link: https://lore.kernel.org/r/20240506114016.30498-9-wsa+renesas@sang-engineering.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/tty/serial/sh-sci.c

index 7a7bc18e47055736442eabef80902ea1bb63c71b..d31b3aca598787525c11dbae3d11fc39955f8b66 100644 (file)
@@ -1324,14 +1324,14 @@ static void sci_dma_rx_complete(void *arg)
        dev_dbg(port->dev, "%s(%d) active cookie %d\n", __func__, port->line,
                s->active_rx);
 
+       hrtimer_cancel(&s->rx_timer);
+
        uart_port_lock_irqsave(port, &flags);
 
        active = sci_dma_rx_find_active(s);
        if (active >= 0)
                count = sci_dma_rx_push(s, s->rx_buf[active], s->buf_len_rx);
 
-       start_hrtimer_us(&s->rx_timer, s->rx_timeout);
-
        if (count)
                tty_flip_buffer_push(&port->state->port);
 
@@ -1354,6 +1354,9 @@ static void sci_dma_rx_complete(void *arg)
        uart_port_unlock_irqrestore(port, flags);
        dev_dbg(port->dev, "%s: cookie %d #%d, new active cookie %d\n",
                __func__, s->cookie_rx[active], active, s->active_rx);
+
+       start_hrtimer_us(&s->rx_timer, s->rx_timeout);
+
        return;
 
 fail: