tech-kern archive

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

Re: zero runtime when negative

On 26 Jun, 2012, at 15:05 , Michael van Elst wrote:
> Most suggestions have to do with selecting a reasonably working emulated
> hardware timecounter. Some suggest VMI, which is a set of paravirtual
> drivers that can access a virtual timecounter provided by vmware.

Ah.  If I'm interpreting that properly what they are recommending is that
the ACPI power management counter be used for the system clock and that
the hardclock() interrupt frequency be set to 100 Hz rather than the 1 kHz Linux
might prefer.  I think the ACPI counter increments at 3.58 MHz, which means
that if it is a 32 bit counter it will take 20 minutes to wrap and the system
should be able to miss a whole lot of clock interrupts without something
happening that is unavoidably bad for the clock.  If what is necessary for
decent timekeeping on a virtual machine is to use a slow counter and not to
expect that clock interrupts will occur on time it should be possible to
make that work.

I guess the next question would be, what hardware clock source is in use
on the machines that see the clock go backwards?

Dennis Ferguson

Home | Main Index | Thread Index | Old Index