[ntp:questions] Re: Is ntp really what I need?

Paul Hilton paul_hilton at somedomain.com
Mon Feb 14 21:20:37 UTC 2005


On Mon, 14 Feb 2005 15:54:47 -0500, Richard B. Gilbert wrote:

> IRQ priority is not the issue!   Code that disables or masks interrupts 
> for extended periods is the usual cause of this sort of thing.  Linux 
> and Windoze are both noted for this behavior.

As I am used to writing drivers for microcontrollers etc. I am pretty
disgusted that anyone would mask interrupts for more than a few
instructions. 1ms (assuming 1kHz) seems a hell of a long time to suspend
interrupts for. Is it the kernel itself (with all it's modules) which is
guilty of this? Is user space even capable of this? It is amazing that the
system is so reliable if it loses any interrupts at all.

As a palliative it seems like the advice to reduce the clock speed to
100Hz may be good, and, if has as been claimed, 2.6.x kernel's timer
interrupt service routined can detect and correct for missed interrupts
might help too.

This started as an irritating error in the clock, it seems to be
indicative of much more serious problems.

Thanks,
Paul Hilton



More information about the questions mailing list