Re: [Qemu-devel] [PATCH 01/11] migration: assert colo instead of check

[Peter Xu <peterx@redhat.com>]

On Wed, Jan 03, 2018 at 09:38:52AM +0100, Juan Quintela wrote:
> Peter Xu <peterx@redhat.com> wrote:
> > When reaching here if we are still "active" it means we must be in colo
> > state.  Assert it instead of check it in if condition.
> 
> I don't think so.
> 
> > Finally I want to use "switch" here rather than lots of complicated if
> > clauses.
> >
> > Signed-off-by: Peter Xu <peterx@redhat.com>
> > ---
> >  migration/migration.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/migration/migration.c b/migration/migration.c
> > index 4de3b551fe..0ee4b4c27c 100644
> > --- a/migration/migration.c
> > +++ b/migration/migration.c
> > @@ -2309,7 +2309,8 @@ static void *migration_thread(void *opaque)
> >          }
> >          runstate_set(RUN_STATE_POSTMIGRATE);
> >      } else {
> > -        if (s->state == MIGRATION_STATUS_ACTIVE && enable_colo) {
> > +        if (s->state == MIGRATION_STATUS_ACTIVE) {
> 
> We want to run this code iff:
> - we are in ACTIVE state
> - we are using colo
> 
> We can be doing a normal migration, with colo compliled in, but not
> enabled, no?

If COLO is not enabled (even if it is compiled in), IMHO we won't
reach this state.  Note that we have this in migration_completion():

    if (!migrate_colo_enabled()) {
        migrate_set_state(&s->state, current_active_state,
                          MIGRATION_STATUS_COMPLETED);
    }

That's where we kept the ACTIVE state, and that should be the only
place. And, this is exactly what this patch is going to do - I want to
remove COLO hacks if possible.

Thanks,

-- 
Peter Xu