[LinuxPPS] http://lkml.org/lkml/2008/4/11/42

[Rodolfo Giometti]

On Sun, Apr 20, 2008 at 10:53:47AM +0200, Bernhard Schiffner wrote:
> Am Samstag, 19. April 2008 18:37 schrieb Rodolfo Giometti:
> 
> 
> > Step 1) Fix up the lowlevel x86-only irq method which improve
> >         resolution so much.
> >
> > 	This step should be easy to implement regarding the timestamp
> > 	issue, a bit complicated regarding the echo function...
> >
> Best to do for the PPS-community.
> But  I doubt you can sell it as PPS to lkml. Better to speak about 
> an "interrupt-tracer" or use any other buzzword.
> 
> BUT:
> PPS is already on the radar of Andrew Morton, so I think its the best to go 
> the recent way as far as we can.
> (And buy Alans tty->ops knowing not to be ideal, but "requested" as a step 
> to pass. What happens later on: let#s be prepared. )

Do you propose to not implement any changes? I'm a bit confused, can
you please explain it better?

> > Step 2) We can completely remove the PPS code from tty serial code
> > 	by using a "generic" client (maybe called pps-generic-irq?)
> > 	which can be used with any IRQ source.
> >
> > 	We can implement something that loaded as follow:
> >
> > 	$ modprobe pps-generic-irq.ko irqs=1,4
> >
> > 	enables two PPS sources linked with IRQs 1 and 4.
> >
> > The first step should improve resolution and the second should remove
> > the problem with serial code.
> 
> No remove (now).
> What is about shared interrupts?

There should be problems at all...

> How does NTP (as primary PPs-user) handle such an situation?

NTP should not know about these changes. :)

> > What do you think about these solutions? :)
> >
> > Rodolfo
> 
> IMHO:
> 1.) Continue on inclusion and adapt Alans tty->ops for the serial client.

The problem is that Alan doesn't want a UART only code and doesn't
want PPS code into DCD irq handler... :'( He preferes doing PPS stuff
into a line discipline (which is not possibile without the
lowlevel-irq patch).

> 2) Improve lowlevel-irq. (but don't post to lkml)

Yes, this should be done but I think we should post it to LKML.

Ciao

-- 

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