Re: Did you know that nfs mount requires accurate time?
On Tue, Feb 02, 1999 at 01:29:09PM +0100, Loic Prylli wrote:
> Oscar Levi writes:
> > This brings a couple of suggestions to mind. First, Loic, is there
> > room on the install disks for hwclock about 32K I think. I'm gonna
> > suggest to the d-boot list that we add an optional 'set time' step to
> > the install procedure for the sake of enabling nfs installs. Also, it
> > would be keen if this time-set could be committed to CMOS RAM. (Is
> > that what the UDBs have?)
> Yes sure I can add hwclock.
As I'm looking into this more, I'm finding that there is another
problem, now. Setting the time fixed an RPC timeout and let the
installed Debian system nfs mount the server. However, the setup
disks still fail to nfs mount. I get a report of sendmsg failure with
a result of 51. The errno's say this is a 'level 2 halted' error.
Whatever that means.
> > Also, it seems like it would be helpful to add a note to the README in
> > the disks-alpha directory about nfs installs requiring accurate time.
> I was not aware of that, do you know where is this restriction needed?
> (I know I used NFS for some time between a machine in year 2018 and
> normal ones).
I think the problem was that I set the time from ARC and that this
translated into something very wrong...in the year 1930. When I get
it working, I'll verify the parameters that make this fail.
> > Note that the ARC time is not the same as is expected by hwclock. I
> > originally set the clock with ARC and found that time was in the year
> > 1930 after booted. I recall that there used to be a time program that
> > compreheded ARC time.
> The "epoch selection" should be done by the kernel, but the code there
> may be wrong or incomplete.
The epoch selection is the choice between ARC and UTC in CMOS?
It seems to me, IMHO, that it would be a good thing to permit (not
require) the user set the time from setup.