I have been reading through timer_interrupt() and related
functions for a bit now, and think I have found a potential
problem.
If a virtual CPU is preempted by the hypervisor while a
process is running on it, it looks like that unlucky
process will get "billed" for CPU time during which the
virtual CPU was not running ...
Am I overlooking something, or is it time to figure out
a way to separate virtual time from wall time so we can
account for the actual CPU time used by processes ?
>From what I can see, this might require a change to the
hypervisor interface, but I am not sure. Ideas ?
--
All Rights Reversed
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel