Re: [PATCH RFC 3/7] migration: multifd_send_kick_main()

[Fabiano Rosas <farosas@suse.de>]

Peter Xu <peterx@redhat.com> writes:

> When a multifd sender thread hit errors, it always needs to kick the main
> thread by kicking all the semaphores that it can be waiting upon.
>
> Provide a helper for it and deduplicate the code.
>
> Signed-off-by: Peter Xu <peterx@redhat.com>
> ---
>  migration/multifd.c | 21 +++++++++++++++------
>  1 file changed, 15 insertions(+), 6 deletions(-)
>
> diff --git a/migration/multifd.c b/migration/multifd.c
> index 4afdd88602..33fb21d0e4 100644
> --- a/migration/multifd.c
> +++ b/migration/multifd.c
> @@ -374,6 +374,18 @@ struct {
>      MultiFDMethods *ops;
>  } *multifd_send_state;
>  
> +/*
> + * The migration thread can wait on either of the two semaphores.  This
> + * function can be used to kick the main thread out of waiting on either of
> + * them.  Should mostly only be called when something wrong happened with
> + * the current multifd send thread.
> + */
> +static void multifd_send_kick_main(MultiFDSendParams *p)
> +{
> +    qemu_sem_post(&p->sem_sync);
> +    qemu_sem_post(&multifd_send_state->channels_ready);
> +}
> +
>  /*
>   * How we use multifd_send_state->pages and channel->pages?
>   *
> @@ -746,8 +758,7 @@ out:
>          assert(local_err);
>          trace_multifd_send_error(p->id);
>          multifd_send_terminate_threads(local_err);
> -        qemu_sem_post(&p->sem_sync);
> -        qemu_sem_post(&multifd_send_state->channels_ready);
> +        multifd_send_kick_main(p);
>          error_free(local_err);
>      }
>  
> @@ -787,8 +798,7 @@ static void multifd_tls_outgoing_handshake(QIOTask *task,
>       * is not created, and then tell who pay attention to me.
>       */
>      p->quit = true;
> -    qemu_sem_post(&multifd_send_state->channels_ready);
> -    qemu_sem_post(&p->sem_sync);
> +    multifd_send_kick_main(p);

There's a bug here in the original code:

It's not really safe to call any of these outside of the channel lock
because multifd_save_cleanup() could execute at the same time and call
qemu_sem_destroy() -> qemu_mutex_destroy(), which can assert because we
might be holding the sem_lock.

It seems the reason we get away with this today is merely due to
timing. A subset of this problem was already encountered here:

[PATCH] migrate/multifd: fix coredump when the multifd thread cleanup
https://lore.kernel.org/r/20230621081826.3203053-1-zhangjianguo18@huawei.com

We could probably release the semaphores for all channels at once inside
multifd_save_cleanup() in the main thread. We'd have a
multifd_send_kick_main() in each channel when it fails and this:

void multifd_save_cleanup(void)
{
    ...   
    for (i = 0; i < migrate_multifd_channels(); i++) {
        MultiFDSendParams *p = &multifd_send_state->params[i];

        if (p->running) {
            qemu_thread_join(&p->thread);
        } else {
            multifd_send_kick_main(p);
        }
    }
    for (i = 0; i < migrate_multifd_channels(); i++) {
        qemu_sem_destroy, etc...
    }
    ...
}