[PATCH 3/7] bsp/tms570: fixed sci driver freeze when closing terminal bug

[Pavel Pisa]

Hello Gedare,

On Sunday 15 of November 2015 14:44:24 Gedare Bloom wrote:
> On Thu, Nov 12, 2015 at 5:11 PM, Pavel Pisa <pisa at cmp.felk.cvut.cz> wrote:
> > From: Premysl Houdek <kom541000 at gmail.com>
> >
> > tms570_sci_interrupt_last_close: do not wait for old data transmit.
> >
> > Signed-off-by: Premysl Houdek <kom541000 at gmail.com>
> > ---
> >  c/src/lib/libbsp/arm/tms570/console/tms570-sci.c | 8 +++++---
> >  1 file changed, 5 insertions(+), 3 deletions(-)
> >
> > diff --git a/c/src/lib/libbsp/arm/tms570/console/tms570-sci.c
> > b/c/src/lib/libbsp/arm/tms570/console/tms570-sci.c index 54c1b69..a799a66
> > 100644
> > --- a/c/src/lib/libbsp/arm/tms570/console/tms570-sci.c
> > +++ b/c/src/lib/libbsp/arm/tms570/console/tms570-sci.c
> > @@ -531,9 +531,11 @@ static void tms570_sci_interrupt_last_close(
> >    tms570_sci_disable_interrupts(ctx);
> >    rtems_termios_device_lock_release(base, &lock_context);
> >
> > -  /* Flush device */
> > -  while ( ( ctx->regs->FLR & TMS570_SCI_FLR_TX_EMPTY ) > 0 ) {
> > -    ;/* Wait until all data has been sent */
> > +  if ( 0 /* for flush on close */ ) {
>
> I don't understand, this is dead code now?

we decided to disable code. The condition to wait for last character
to be completely transmitted has been negated. This problem has been
corrected. Code is invoked only in few cases because console is usually
permanently open that is why we did not catch a problem in previous
version and it has not been catch even during some test with the second
UART as well. Attempt to wait for the last character is  based on logic
found in other serial interface drivers.

But even that logic was right after correction, ẅe consider
busy loop as bad option and required only seldom by applications.

The code has been disabled for now. There seem to be no option
to enable interrupt on TX EMPTY event in TMS570 UART hardware.
This means that some form of busy/poll loop is required.

We update patch if your and others prefer to delete it
or implement more complete solution. I would prefer poll
loop with code like this for the second option
  
  rtems_interval tw = rtems_clock_get_ticks_per_second();
  tw = tw * 10 / baudrate + 1;
  while ( ( ctx->regs->FLR & TMS570_SCI_FLR_TX_EMPTY ) == 0 ) {
    rtems_task_wake_after(tw);
  }

> > +    /* Flush device */
> > +    while ( ( ctx->regs->FLR & TMS570_SCI_FLR_TX_EMPTY ) == 0 ) {
> > +      ;/* Wait until all data has been sent */
> > +    }
> >    }
> >
> >    /* uninstall ISR */
> > --
> > 1.9.1
> >
> > _______________________________________________
> > devel mailing list
> > devel at rtems.org
> > http://lists.rtems.org/mailman/listinfo/devel