Hello, ATM chrony enables phc and pps by default. However they are useful for limited cases where user have a special hardware time source (atomic clock, gps clock and so on). Since most users apparently don't have such a hardware, I propose to not enable these options by default, this way users will benefit from a reduced attack surface, smaller memory footprint and so on. Another note, pps USE flag description is wrong: <flag name='pps'>Support for the Linux Real Time Clock (RTC) interface</flag> while pps is "PPS (Pulse Per Second) is a special pulse provided by some GPS antennae." (from CONFIG_PPS description).
What attack surface?
I like to stay close to upstream. So their default configure options are our default configure options, except where they would clash with Gentoo specific requirements. I don't think these two options should be disabled just because they increase the install size. As for the attack surface, that would require a malicious RTC or PPS driver in the kernel or having those open to unprivileged user access. chrony's attack surface would in those cases be the least of your concerns.
(In reply to Jeroen Roovers from comment #2) > As for the attack surface, that would require a malicious RTC or PPS driver PCH and PPS, that is.
PHC...
(In reply to Jeroen Roovers from comment #2) > I like to stay close to upstream. So their default configure options are our > default configure options, except where they would clash with Gentoo > specific requirements. This is understandable, but questionable approach. Upstream is usually targeted on the audience unable/unwilling to rebuild from sources, thus upstream tends to include as functionality as possible except for questionable stuff. > As for the attack surface, that would require a malicious RTC or PPS driver > in the kernel or having those open to unprivileged user access. chrony's > attack surface would in those cases be the least of your concerns. Not really, just some bug in auxiliary code may be triggered. Less code => less bugs, that's simple and works statistically quite well (of course there are some exceptions, e.g. hardening code).
