do_gettimeofday monotony?

do_gettimeofday monotony?

[bert hubert]

Is do_gettimeofday supposed to be monotonous? I'm seeing time go backward by
tiny amounts, and then progressing.

I'm using do_gettimeofday on a single processor, CONFIG_PREEMPT_NONE=y, and
saving stuff from generic_make_request - see http://ds9a.nl/diskstat for the
source. 2.6.13-rc3-mm1, HZ=250.

I'll try to figure out how much it is going back, and it is some kind of
magic interval.

Is there another monotonous clock in the kernel that doesn't wrap (all
that often)? Doesn't really need to be wall time.

Thanks!

-- 
http://www.PowerDNS.com      Open source, database driven DNS Software 
http://netherlabs.nl              Open and Closed source services

Re: do_gettimeofday monotony?

[Christopher Friesen]

bert hubert wrote:
> Is do_gettimeofday supposed to be monotonous?

Nope.

 > I'm seeing time go backward by tiny amounts, and then progressing.

Are you running NTP?  Corrections could cause this.

> Is there another monotonous clock in the kernel that doesn't wrap (all
> that often)? Doesn't really need to be wall time.

clock_gettime()

Chris

Re: do_gettimeofday monotony?

[bert hubert]

On Mon, Jul 25, 2005 at 11:12:25AM -0400, Christopher Friesen wrote:
> >Is do_gettimeofday supposed to be monotonous?
> 
> Nope.
> 
> > I'm seeing time go backward by tiny amounts, and then progressing.
> 
> Are you running NTP?  Corrections could cause this.

No. I am running a machine which often changes it clockspeed though. The bad
news is that I don't appear to be able to reproduce this, there must be
something specific that causes it. But then again, it wasn't a large effect.

> >Is there another monotonous clock in the kernel that doesn't wrap (all
> >that often)? Doesn't really need to be wall time.
> 
> clock_gettime()

Thank you.

-- 
http://www.PowerDNS.com      Open source, database driven DNS Software 
http://netherlabs.nl              Open and Closed source services