Re: [PATCH v3 1/1] yank: Unregister function when using TLS migration

[Leonardo Brás]

On Tue, 2021-06-01 at 10:50 -0400, Peter Xu wrote:
> On Tue, Jun 01, 2021 at 02:40:31AM -0300, Leonardo Bras wrote:
> > [...]
> >  inside migration_channel_connect() and
> > migration_channel_process_incoming() move yank_register_function()
> > so
> > it only runs once on a TLS migration.
> 
> Slightly inaccurate I guess, as it was run once too before this
> patch, but not
> paired for tls?

Yeah, I think I got confused over this:
I remembered having multiple instances of yank_register_function()
during TLS migration on qemu.org/master, but after re-testing it, I
think it was in fact something to do with V1 and V2.

So, yes, that's correct, in qemu.org/master there was a single
yank_register_function() and no pairing yank_unregister_function()
during TLS migration. 

> 
> IIUC when we call the 2nd time at below chunk we won't register
> again:
> 
>         if (object_dynamic_cast(OBJECT(ioc),
> TYPE_QIO_CHANNEL_SOCKET)) {
>             yank_register_function(MIGRATION_YANK_INSTANCE,
>                                    migration_yank_iochannel,
>                                    QIO_CHANNEL(ioc));
>         }
> 
> Because the 2nd call will be TYPE_QIO_CHANNEL_TLS, so
> object_dynamic_cast()
> will return with a failure, I think (note, TYPE_QIO_CHANNEL_TLS's
> parent is
> TYPE_QIO_CHANNEL, not TYPE_QIO_CHANNEL_SOCKET).
> 
> > 
> > Fixes: b5eea99ec2f ("migration: Add yank feature", 2021-01-13)
> > Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1964326
> > Signed-off-by: Leonardo Bras <leobras.c@gmail.com>
> 
> Reviewed-by: Peter Xu <peterx@redhat.com>
> 
> Thanks Leo!

Thank you Peter!