Re: [netbsd-5] machine losing time by the hour

[Anne Bennett <anne%porcupine.montreal.qc.ca@localhost>]

Hauke Fath <hf%spg.tu-darmstadt.de@localhost> reports:

> since I have upgraded a few servers to fresh netbsd-5 snapshots, one of the
> machines (a P4/3 GHz, Intel server board, GENERIC kernel) started to lose
> time like a sieve. The clock seems to run half speed, ntpd(8) is unable to
> reign in the box, and does not produce any diagnostics, either.
> 
> I suspected some weird HW failure, but rolling back the machine to a
> 2009-09-21 kernel fixes the issue.
> 
> Does this regression ring a bell for anyone? Should I send-pr?

It is similar to a temporary condition I saw a few days ago right
after booting back into a 5.0.1 release kernel after failing to
boot from a netbsd-5 kernel (crashed with i195drm problem during
boot).  Rebooting again in the 5.0.1 release kernel fixed the problem.
It happened twice, and both times, the clock slowed down (I would
issue "date" repeatedly and not see the time advancing for many
seconds), and also mouse selections became unreliable.

I described the problem in port-i386/42966, but it was tangential to
the main issue (crashes on boot due to i915 driver and DIAGNOSTICS),
which is now fixed in -current.

The tentative conclusion was that the faulty netbsd-5 kernel had
left the hardware in a funny state, but we didn't really investigate
that aspect.  I'm now running the fixed netbsd-5 kernel, and I see
no such problems.

The machine is an HP Pavilion a1620n, which has a dual core CPU of
type "Intel 686-class, 3001MHz, id 0xf65".

Hope that helps...


Anne.
-- 
Ms. Anne Bennett, as a private citizen:  anne%porcupine.montreal.qc.ca@localhost
Also reachable more officially at work:  anne%encs.concordia.ca@localhost