Re: [PATCH 2/4] multifd: Make sure that we don't do any IO after an error

[Juan Quintela]

"Dr. David Alan Gilbert" <address@hidden> wrote:
> * Juan Quintela (address@hidden) wrote:
>> Signed-off-by: Juan Quintela <address@hidden>
>
> Reviewed-by: Dr. David Alan Gilbert <address@hidden>
>
> I wonder if the fflush's should happen anyway?

No, that is the problem that I am really trying to fix.  We tried to do
a write() after we knew that we have closed it (due to error or
cancel).  And we get a nice error message on the screen:

Unable to write to socket()

And everybody gets scared about that message.  When we really know that
we don't want it.

In an ideal world, we would just remove ->last_error() and make every
function return errors and check return codes.  But this is qemu.  This
is migration.  And we don't do it.  Sniff.