[LinuxPPS] multiple access to /dev/ppsN entries

Bernhard Schiffner bernhard at schiffner-limbach.de
Thu Mar 18 10:52:21 CET 2010


On Thursday, 18. March 2010 10:08:45 Miroslav Lichvar wrote:
> On Wed, Mar 17, 2010 at 08:55:15PM +0100, Bernhard Schiffner wrote:
> > Again:
> > 1.)  An "Unixer" simply expects something useable from reading a
> > character device.
> > 2.) The RFC2783 does neither prohibit nor force a human readable readout
> > of /dev/pps.
> > 3.) Mar 2009 were some related discussions about (IIRC) /dev/time(?) |
> > Plan9 | BSD etc. There was no conclusion about a standardized format of
> > timestamps. I don't know the state of the onion today.
> > 4.) Mar 2009 were other priorities to work on first (inclusion into
> > mainline, hick hack about ldisk etc.)
> > 5.) I think it's a good time to discuss this topic again.
> > 6.) This thread is a base to do so.
> > 7.) Last and very least George and me want this feature in one or an
> > other form implemented.  A patch exists.
> > 
> > What's your POW?
> 
> I'm not involved in the development, but having a file descriptor that
> can be used in select() (so the application doesn't need to do any
> polling) seems very useful to me.

+=1

int select(int nfds, fd_set *readfds, fd_set *writefds,
           fd_set *exceptfds, struct timeval *timeout);

Georges patch makes a read() possible every time.
So  a  select(...) on  *readfds makes  no sense. The only way is to "fill" 
*exceptfds by the driver when receiving a new timestamp.

In some private mail  with Ben Gardiner it shows up, that something like this 
is his favorite way to connect to LinuxPPS functionality without using the 
RFC2783 API.

Now: I'am not a programmer.
Do you know someone to find a solution therefore?
(Until now it's only my very limited suggestion.)


Bernhard



More information about the LinuxPPS mailing list