[ntp:hackers] Re: More questions on the ATOM driver.

David L. Mills mills at udel.edu
Thu Mar 23 16:48:39 UTC 2006


Don't get me going about Linux. The kernel mods for all systems use 
essentially the same code, which makes it easier to spot and clear bugs. 
But, it doesn't really matter how the functions are implemented, just 
that they are the same exact functions. The clock discipline and PPS 
functions need to be mathematically exact, not approximate. I strongly 
suspect the various Linux patches will be infected with other ideas.


Reg Clemens wrote:

>>I think your mission to stand down the kernel PPS is probably moot. Do 
>>you expect HP, Sun, Cisco and FreeBSD to remove the support now that 
>>it's been in their products for over a decade?
> No, but then this PPS code has never made it into the Linux kernel.
> Ulrich had a patch for the 2.4.x series of kernels that included both the
> PPS capture code and the PPS kernel code, but it was never submitted to
> become part of the official kernel tree.
> Since then Ulrich has had a much smaller patch for the 2.6.x kernel tree
> that just has the PPS capture code in it, and not the 'enable kernel' stuff.
> As noted on this list a month or so ago, Rodolfo Giometti has a similar
> patch, using more modern methods to communicate from the user to the kernel.
> So for Linux, and what we consider modern, fast, machines, it would seem
> to me, that this new patch could be enough.  Thats where Im really heading.
> There's no question of removing it where it currently exists, just the
> question of NOT writing and installing it if it doesnt exist.

More information about the hackers mailing list