Paolo Bonzini <pbonz...@redhat.com> writes:
> Il 21/02/2013 18:25, Anthony Liguori ha scritto:
>>> >
>>> > What would be more problematic is the chardev flow control patches,
>>> > which use the glib main loop directly. I don't recall your KVM forum
>>> > presentation---did you need RT prioritization of the serial port too?
>> It uses GSources which don't need a full glib main loop. We just need
>> to be able to support glib event dispatch from whatever our main loop
>> is.
>
> The lock in g_main_context_{prepare,query,check,dispatch} is not
> RT-friendly. We could have contention on that lock between the iothread
> and the VCPU thread.
Since I assume we'll have multiple I/O threads by then, I also assume
that we'll run RT events on one or more dedicated I/O thread that only
handles RT events.
In terms of attaching events to I/O threads, it's pretty simple to delay
the GSource attach to a bottom half on a non-RT I/O thread if the VCPU
needs to add a glib based event.
Regards,
Anthony Liguori
>
> Paolo
>
>> I would assume that we'd treat any glib event with the same priority if
>> we ever had a RT-iothread.
