>> I don't really like updating the last_start in burn_budget().
>>
>> last_start is the recent starting time of the VCPU. It is to record
>> the starting time of a currently running VCPU. So it should be
>> updated
>> only when the VCPU is scheduled.
>>
> Yes, the name suggest it is what you write
On Wed, 2016-10-26 at 12:08 -0400, Meng Xu wrote:
> On Wed, Oct 26, 2016 at 10:43 AM, Dario Faggioli
> wrote:
> >
> > On Fri, 2016-10-21 at 22:11 -0400, Meng Xu wrote:
> > >
> > > Bug scenario:
> > > While a VCPU is running on a core, it may misses its deadline.
> > >
> > May be useful to menti
On Wed, Oct 26, 2016 at 10:43 AM, Dario Faggioli
wrote:
> On Fri, 2016-10-21 at 22:11 -0400, Meng Xu wrote:
>> Bug scenario:
>> While a VCPU is running on a core, it may misses its deadline.
>>
> May be useful to mention why (at least the most common reasons, like
> overhead and/or system being ov
On Fri, 2016-10-21 at 22:11 -0400, Meng Xu wrote:
> Bug scenario:
> While a VCPU is running on a core, it may misses its deadline.
>
May be useful to mention why (at least the most common reasons, like
overhead and/or system being overloaded... are there others?).
> Then repl_timer_handler() will
On Fri, Oct 21, 2016 at 10:11 PM, Meng Xu wrote:
> Bug scenario:
> While a VCPU is running on a core, it may misses its deadline.
> Then repl_timer_handler() will update the VCPU period and deadline.
> The VCPU may have high priority with the new deadline and repl_timer_handler()
> decides to keep
Bug scenario:
While a VCPU is running on a core, it may misses its deadline.
Then repl_timer_handler() will update the VCPU period and deadline.
The VCPU may have high priority with the new deadline and repl_timer_handler()
decides to keep the VCPU running on the core, but with new period and deadl
