Re: AGAIN: Re: gettimeofday clock jump bug (on AMD 756)

[Karim Yaghmour]

"Ph. Marek" wrote:
> The difference is both times about 4295 seconds - so I think that the problem
> has something to do with 2^32 microseconds.

This 2^32 jump is consistent with the jumps I saw.

I was discussing this issue with Richard Moore (IBM) a couple of weeks
ago and he mentionned that they had seen a similar problem with OS/2.
In that case, the problems were due to the time it took to read some
timer register. In other words, there may be a short time-window where
the values available are valid and nothing is garanteed if we exceed this
time-window. (This is second hand and I may have misunderstood a couple
of details, so I attached Richard so he can confirm/deny what I'm saying
here.)

Linux reads the PIT (8253) every 10ms on a PC, so the question is:
anyone know about 8253 reading deadlines ?

Karim

===================================================
                 Karim Yaghmour
               [e-mail]
      Embedded and Real-Time Linux Expert
===================================================