Subject: Re: CVS commit: src/sys/arch/cesfic
To: None <M.Drochner@fz-juelich.de>
From: Garrett D'Amore <garrett_damore@tadpole.com>
List: source-changes
Date: 09/08/2006 07:47:04
Matthias Drochner wrote:
> [systems w/o RTC]
>
> garrett_damore@tadpole.com said:
>   
>> These systems will now, by default, get a warning on boot that "no TOD
>> clock present" and "CHECK AND RESET THE DATE!" 
>>     
>
> Systems with NFS root (the only choice on cesfic) use the last
> access time of the root directory immediately after mounting root
> as initial time.
> This is arguably more accurate than a typical RTC... such a warning
> is not appropriate at all.
>
> best regards
> Matthias
>
>   

I disagree with you, although I will bow to the will of the group if
that is the consensus.  But let me first make my points:

a) we have no way of knowing how long the system has been powered off. 
if you shut the system off for two weeks, then the root filesystem is
probably 2 weeks old!

b) on such systems, the admin probably should be running
NTP/ntpdate/rdate.  If they aren't then something is wrong

c) finally, the warning is accurate, and lets admins know that this
behavior (using the filesystem date) is taking place.  It might not
otherwise be apparent (how many people know whether their system has a
functional RTC or not?), and the warning makes it clear.

d) if the admin knows that he's using rtc/ntpdate/ntp or the system was
only momentarily rebooted, then he can ignore the warning.

e) this warning is also used if the TOD driver isn't configured or
doesn't attach  (on some systems this behavior used to cause a panic. 
now it causes the above behavior uniformly)

f) I'd long since adopted this policy on the AR531x port and alchemy
ports.  The ar531x parts do not have an RTC.  Only newer alchemy parts
have an RTC that is battery backed (the RTC keeps time across reboots,
but only if power is not lost, on other boards.)

Anyway, I think the warning does more good than harm.  However, I am
willing to allow that maybe some port masters might strongly object to
it, and want it removed.  I think that such a decision should be made
very carefully before it is undertaken, however.  I think there is a
strong benefit in having all systems behave similarly here.

-- 
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
http://www.tadpolecomputer.com/
Phone: 951 325-2134  Fax: 951 325-2191