On 01/22/2016 08:28 AM, John Ogness wrote:
> If DMA is active during a shutdown, a delayed restore of the
> registers may be pending. The restore must be performed after
> the DMA is stopped, otherwise the delayed restore remains
> pending and will fire upon the first DMA TX complete of a
> totally different serial session.
>
> Signed-off-by: John Ogness <john.ogn...@linutronix.de>
> ---
> This patch was originally posted and discussed here:
> http://lkml.kernel.org/r/87egjp2r4a....@linutronix.de
>
> The patch was not accepted because Peter expressed dissatisfaction with
> the entire deferred implementation. (This patch fixes a bug in that
> implementation.) However, no alternative solution was determined and the
> bug continues to exist.
Yeah, I made that comment a while ago and still no one's fixed that.
What gives?
Regards,
Peter Hurley
> Although we can certainly continue to debate about an alternate solution
> to the deferred implementation, I think it makes sense to at least fix
> the existing bug until an alternative exists.
>
> The bug can be easily produced with a `kill -9` on a process transmitting
> on the serial port.
>
> patch against next-20160122
>
> drivers/tty/serial/8250/8250_omap.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/tty/serial/8250/8250_omap.c
> b/drivers/tty/serial/8250/8250_omap.c
> index a2c0734..0ffec67 100644
> --- a/drivers/tty/serial/8250/8250_omap.c
> +++ b/drivers/tty/serial/8250/8250_omap.c
> @@ -664,9 +664,15 @@ static void omap_8250_shutdown(struct uart_port *port)
> up->ier = 0;
> serial_out(up, UART_IER, 0);
>
> - if (up->dma)
> + if (up->dma) {
> serial8250_release_dma(up);
>
> + if (priv->delayed_restore) {
> + priv->delayed_restore = 0;
> + omap8250_restore_regs(up);
> + }
> + }
> +
> /*
> * Disable break condition and FIFOs
> */
>
