Messages in this thread | | | Date | Thu, 3 Apr 2008 08:28:25 +0200 (CEST) | From | Thomas Gleixner <> | Subject | Re: gettimeofday() jumping into the future |
| |
On Thu, 3 Apr 2008, Karsten Wiese wrote: > > To prevent this TSC specific wreckage we need to compare the TSC value > > against the reference value and return the latter when it is larger > > than the actual TSC value. > > Last paragraph IMO should be: > To prevent this TSC specific wreckage we need to substract the > reference value from the TSC value, interpret the result as > signed. If the interpreted result is negative, return the reference > value, else the TSC Value. > Your patch misses the corner case where TSC value is < reference _and_ valid > at TSC wrap around.
In about 200 years on a 3Ghz box.
> The access to the reference value needs a (the xtime ?) lock on 32bit, no?
The readout is under xtime lock anyway.
Thanks,
tglx
| |