Re: [Qemu-devel] [PATCH] do not call monitor_resume() from migrate_fd_put_buffer() error path

[Luiz Capitulino]

On Wed,  3 Aug 2011 18:51:44 +0400
Michael Tokarev <address@hidden> wrote:

> If we do, it results in double monitor_resume() (second being called
> from migrate_fd_cleanup() anyway) and monitor suspend count becoming
> negative.

Are you hitting an specific issue or did you find this by code inspection?

IIRC, I asked Marcelo to add the monitor_resume() call in the fix for
e447b1a603 because the monitor wasn't being resumed in some cases. Don't
remember which though, do you Marcelo?

I see two possibilities here:

 1. After e447b1a603 there was some change that made the monitor_resume() call
    in migrate_fd_put_buffer() unnecessary

 2. We're calling it in the wrong place

Taking a quick look at the code I see that migrate_fd_cleanup() doesn't
seem to be called when qemu_savevm_state_iterate() fails, for example.

> 
> Signed-Off-By: Michael Tokarev <address@hidden>
> Reviewed-By: Jan Kiszka <address@hidden>
> ---
>  migration.c |    3 ---
>  1 files changed, 0 insertions(+), 3 deletions(-)
> 
> diff --git a/migration.c b/migration.c
> index 2a15b98..7ca883f 100644
> --- a/migration.c
> +++ b/migration.c
> @@ -330,9 +330,6 @@ ssize_t migrate_fd_put_buffer(void *opaque, const void 
> *data, size_t size)
>      if (ret == -EAGAIN) {
>          qemu_set_fd_handler2(s->fd, NULL, NULL, migrate_fd_put_notify, s);
>      } else if (ret < 0) {
> -        if (s->mon) {
> -            monitor_resume(s->mon);
> -        }
>          s->state = MIG_STATE_ERROR;
>          notifier_list_notify(&migration_state_notifiers, NULL);
>      }