Should the HAVE_KERNEL_PLL conditional be abolished?

Hal Murray hmurray at
Sun Dec 3 20:21:19 UTC 2017

> Should the HAVE_KERNEL_PLL conditional be abolished?  Or. to put it
> equivalently, do we have any development targets that lack ntp_adjtime()/
> adjtimex()? 

That symbol is not defined when I build on OpenBSD.

I'd be happy to clean up that area, but I think it is more complicated than 
you have indicated.

There are 2 RFCs 
  RFC1589: Kernel Model for Precision Timekeeping
  RFC2783: Pulse-Per-Second API

There are actually two interesting parts to that tangle.

The first is getting timing data for a pulse.  PPS for NTP is the main use, 
but it can also be used for general hacking.  (I use it to monitor the power 
line frequency.)

The second is a PLL inside the kernel that gets switched on if you set a bit. 
 That works significantly better than whatever ntpd does.  I think we should 
be able to fix this.  It's not a trivial fix.  It may require implementing 
the wake-on-data-ready option which doesn't seem to be implemented.  (but 
maybe I didn't look in the right place)

The "HAVE_KERNEL_PLL" is used for both of them even though there is nothing 
PLL related to the first usage.

These are my opinions.  I hate spam.

More information about the devel mailing list