Re: [Qemu-devel] [PATCH v2 1/1] char/serial: Fix emptyness handling

[Peter Crosthwaite]

On Thu, Feb 27, 2014 at 12:48 PM, Don Slutz <address@hidden> wrote:
> The commit 88c1ee73d3231c74ff90bcfc084a7589670ec244
> char/serial: Fix emptyness check
>
> Still causes extra NULL byte(s) to be sent.
>
> So if the fifo is empty, do not send an extra NULL byte.
> Do full state change on fifo8_is_empty.
>
> Signed-off-by: Don Slutz <address@hidden>
> ---
> v1 to v2:  Do all the state changes that would have been done sending the 
> NULL byte.
>
>  hw/char/serial.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/hw/char/serial.c b/hw/char/serial.c
> index 6d3b5af..7af3c1b 100644
> --- a/hw/char/serial.c
> +++ b/hw/char/serial.c
> @@ -225,8 +225,13 @@ static gboolean serial_xmit(GIOChannel *chan, 
> GIOCondition cond, void *opaque)
>
>      if (s->tsr_retry <= 0) {
>          if (s->fcr & UART_FCR_FE) {
> -            s->tsr = fifo8_is_empty(&s->xmit_fifo) ?
> -                        0 : fifo8_pop(&s->xmit_fifo);
> +            if (fifo8_is_empty(&s->xmit_fifo)) {
> +                s->lsr |= UART_LSR_THRE | UART_LSR_TEMT;
> +                s->thr_ipending = 1;
> +                serial_update_irq(s);
> +                return FALSE;
> +            }

This is copy paste of the UART_LSR_THRE handler code below. The
implementation is now somewhat inconsistent with non-fifo mode with
the mixed stage handling of empty fifo/hold-reg. Perhaps you could:

if (s->fcr & UART_FCR_FE && fifo8_is_empty(&s->xmit_fifo)) {
   s->lsr |= UART_LSR_THRE;
}

after a a successful transmit? Then let the existing check of
UART_LSR_THRE catch for your irq updates etc.

> +            s->tsr = fifo8_pop(&s->xmit_fifo);
>              if (!s->xmit_fifo.num) {
>                  s->lsr |= UART_LSR_THRE;
>              }

I think this is this now dead code. I think (!s->xmit_fifo.num) is the
same condition as fifo8_is_empty().

Sorry about the delayed response. Thanks for the patch.

Regards,
Peter

> --
> 1.8.4
>
>