Re: LinuxPPS low-level IRQs timestamps & ldisc

[Rodolfo Giometti <giometti@enneenne.com>]

On Mon, Jun 02, 2008 at 03:02:57PM +0100, Alan Cox wrote:
> > It could be a bit better... I did as above since I supposed that you
> > wished all PPS code should be removed from serial port code.
> 
> Ideally. But if it is genuinely the case that the serial port IRQ handler
> in some cases needs to do
> 
> my_interrupt() {
> 	get_timestamp()
> 	frob_with_hardware()
> 	ld->dcd_change(blah, timestamp)
> }
> 
> then that is still fairly clean and more importantly actually appears to
> work. I'd avoid all the ifdefs with this in the serial drivers that need
> more accuracy:
> 
> {
> 	struct timespec ts;
> 	if (ld->dcd_change)
> 		getnstimeofday(&ts);
> 	existing tty stuff
> 	if (ld->dcd_change)
> 		ld->dcd_change(tty, status, &ts);
> }

I prefere avoid the if clause for getnstimeofday() since each
instruction delay may decrease time precision, so:

{
      struct timespec ts;
      getnstimeofday(&ts);
      existing tty stuff
      if (ld->dcd_change)
              ld->dcd_change(tty, status, &ts);
}

In the wrost case the timestamp is not used.

> And in ld->dcd_change do
> 
> 	struct timespec myts;
> 	/* Caller passed NULL meaning 'do your own timestamp' */
> 	if (ts == NULL) {
> 		ts = &myts;
> 		getnstimeofday(&myts);
> 	}

Ok.

> so for uart that would
> 
> 	uart_dcd_change(struct uart_port *port, unsigned int status,
> 					struct timespec *ts) {
> 	{
> 		...
> 	}
> 
> passing the timestamp from the ttys own IRQ handler

Great!

Thanks,

Rodolfo

-- 

GNU/Linux Solutions                  e-mail:    giometti@enneenne.com
Linux Device Driver                             giometti@linux.it
Embedded Systems                     phone:	+39 349 2432127
UNIX programming                     skype:     rodolfo.giometti