Re: unreviewed m68k patches

[Bernardo Innocenti <bernie at develer dot com>]

Richard Henderson wrote:

> > - Add __attribute__((interrupt_handler))
> > (Not needed for uClinux, useful for embedded developers)
> > http://gcc.gnu.org/ml/gcc-patches/2003-09/msg01262.html
>
> I guess this is ok.  It'd be nice if you'd generate a fake
> exception frame when these functions are called directly,
> but I suppose no one actually does that.  ;-)

Even in this era of strict compile time checks, whoever
dares calling an interrupt handler in interrupt handler
as it was a normal function deserves what they get :-)

Well, actually I've seen this pattern used in some
embedded programs:

void serial_tx_irq()
{
  if (buffer_not_empty)
     send_one_byte;
  else
     sending = false;
}

void serial_putchar(char c)
{
  buffer_push(c);

  disable_irqs;
  if (!sending)
  {
      sending = true;
      serial_tx_irq();
  }
  enable_irqs;
}

If serial_tx_irq() is given the interrupt_handler attribute,
it's not going to work any more. One would have to resort
to a linux-ism such implementing serial_tx_irq() in terms
of a plain function __serial_tx_irq() and make that
always_inline for efficiency.

Bleah, embedded code sometimes really sucks ;-)

--
 // Bernardo Innocenti - Develer S.r.l., R&D dept.
\X/  http://www.develer.com/

Please don't send Word attachments - http://www.gnu.org/philosophy/no-word-attachments.html