[ntp:questions] new driver development

Dave Hart hart at ntp.org
Tue Mar 29 10:08:15 UTC 2011

On Tue, Mar 29, 2011 at 7:41 UTC, Martin Burnicki
<martin.burnicki at meinberg.de> wrote:
> Whether addressing of either mode is done by separate driver numbers (which
> I still find a good idea in this case, since SHM is a generic driver), or
> by a mode number for the existing SHM driver, is a totally different
> question.

I'm leaning towards using the existing shared memory layout and
protocol for units 0-3, which until relatively recently were the only
possibilities, and POSIX named shared memory with new, safer layout
and protocol for units 4 and up.  gpsd today only provides 0-3.  0 & 1
have root-only permission, while 2 & 3 are world-writable.  I'm also
tending towards ntpd never creating the POSIX named shared memory nor
relying on its presence to configure the clock, but attempting to open
it read-only if it's not already on each poll, and close it if there
are no updates for a timeout period.

Dave Hart

More information about the questions mailing list