Arjan van de Ven wrote:
Even ntp should be better off with this more finegrained data, but I can believe it'll need to get used to it for some time ;)
Can anyone report on whether the lost tick interrupt problem at 1000Hz interrupt rate been firmly resolved on slower machines and machines wih the CPU speed stepped down? I haven't looked at the interrupt handling code in some time. The net effect of a lost interrupt was the appearance that the system time stood still for one clock interval out of the two that had just passed. This caused significant variations in system time. Ntp was forced to accelerate the clock until it caught up with true time and then deaccelerate to avoid overshoot. This was repeated with every lost interrupt resulting in system time that observed a "sawtooth" path relative to true time. -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
