Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: 4.99.45 vs. 4.99.36: bad timekeeping on sparc64

Hi Patrick !

Patrick Welche wrote:
On Fri, Dec 28, 2007 at 11:29:10PM -0500, Rafal Boni wrote:
I'm running two sparc64 boxes on different versions of -current -- one a
V100 running 4.99.36 (updated rarely since it's my mail server, network
storage for the PCs, etc.) and a Netra T1 running 4.99.45.  Both point
at the same NTP server (an older NetBSD/i386 box that's peered to some
external NTP sources)...

The older box (running 4.99.36) keeps great time.. the newer box,
however, loses so much time it's about 40 seconds off now, and it hasn't
been up more than 1.5 days.

So, at least on sparc64, the CVS HEAD seems to keep pretty abysmally bad
time.  I haven't yet dug into any details but thought I'd send a query /
heads-up here.

Just a "me too" on i386 - just noticed 2 days ago, but haven't narrowed
it down.
Could you try "sysctl -w kern.timecounter.hardware=ACPI-Fast" and see if that makes a difference.

machdep.cpu_brand = AMD Sempron(TM) 2200+
kern.timecounter.choice = clockinterrupt(q=0, f=100 Hz) TSC(q=-100, 
f=1800183380 Hz) ACPI-Fast(q=1000, f=3579545 Hz) lapic(q=2100, f=400014291 Hz) 
i8254(q=100, f=1193182 Hz) dummy(q=-1000000, f=1000000 Hz)
kern.timecounter.hardware = lapic
kern.timecounter.timestepwarnings = 0



Home | Main Index | Thread Index | Old Index