Fabio M. Di Nitto napsal(a):
> Only one question inline
>
> On 4/5/2013 10:36 AM, Jan Friesse wrote:
>
>> +static void timer_function_scheduler_timeout (void *data)
>> +{
>> + struct scheduler_pause_timeout_data *timeout_data = (struct
>> scheduler_pause_timeout_data *)data;
>> + unsigned long long tv_current;
>> + unsigned long long tv_diff;
>> +
>> + tv_current = timerlist_nano_current_get ();
>
> are we sure the value returned from this function is increased regularly
> even if the VM is "slowing down" from outside (host scheduling)?
Nope, at least not generally. I was trying qemu-kvm and here it works
(it's really increased as expected, it's possible to sigstop qemu-kvm
process, then sigcont and diff is correct, it's also possible to
overload host (while true do true) and it works). But I'm really unsure
how other hypervisors behaves.
>
> Otherwise the patch looks good to me (nitpick a couple of extra \n)
>
> Fabio
> _______________________________________________
> discuss mailing list
> [email protected]
> http://lists.corosync.org/mailman/listinfo/discuss
_______________________________________________
discuss mailing list
[email protected]
http://lists.corosync.org/mailman/listinfo/discuss
