[systemd-devel] systemd 28 time handling changes "breaks" ntpdate and date ?

[Kay Sievers]

On Apr 27, 2012 2:11 AM, "Colin Guthrie" <gmane at colin.guthr.ie> wrote:
>
> 'Twas brillig, and Sam Morris at 25/04/12 23:41 did gyre and gimble:
> > On Wed, 25 Apr 2012 11:28:14 -0400, Martin Langhoff wrote:
> >
> >> Apologies if this has been discussed -- a search through the archives
> >> didn't find anything.
> >>
> >> Reading the announcement of systemd v28...
> >>
> >> "At shutdown we no longer invoke "hwclock --systohc", i.e. do not write
> >> the system clock back to the RTC. Why? In general there's not really a
> >> reason to assume that the system clock was anymore correct than the RTC
> >> so it's probably a good idea to leave the RTC untouched."
> >
> > Hm, every time I reboot I'm prompted to run fsck because the last mount
> > time is in the future. Is this the reason?
>
> Yes, this is the reason.

The real reason is that the extfs tools just make silly assumptions. You
should patch it away or disable that nonsense in the extfs config file,
which is what Fedora does.

> But that said there should be an automatic sync if you run chrony or
> ntpd every 12s or so at the kernel level.

Every 11 minutes. The kernel writes to the rtc when ntp puts the kernel
into that mode.

There is only one piece missing, the kernel syncs not over deltas larger
than 15 or 30 minutes, even when ntp sets the clock to that. That can get
into our way when daylight saving time changes and the rtc is in localtime.
It should be fixed in the kernel, by adding a mode that says
full-sync-if-in-11-minute mode.

Kay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/systemd-devel/attachments/20120427/8a5ff9b4/attachment.htm>