Re: [virtio-dev] Clarification of VIRTIO_F_EVENT_IDX driver conditions.

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Jun 01, 2022 at 01:03:27PM -0700, Patrick Mosca wrote:
> We ran into an edge case in the VIRTIO_F_EVENT_IDX feature where a notification could be missed if there is a "streak" in skipped notifications. If the difference between the event index and available index wraps before a notification is needed, there exists a case where a notification must be sent but is not sent.
> 
> 
> Signed-off-by: Patrick Mosca <pmosca@apple.com>
> Signed-off-by: Cameron Esfahani <dirty@apple.com>
> ---
>  split-ring.tex | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/split-ring.tex b/split-ring.tex
> index bfef62d..9650e5e 100644
> --- a/split-ring.tex
> +++ b/split-ring.tex
> @@ -530,6 +530,9 @@ \subsection{Available Buffer Notification Suppression}\label{sec:Basic Facilitie
>          \item If the \field{idx} field in the available ring (which determined
>            where that descriptor index was placed) was equal to
>            \field{avail_event}, the driver MUST send a notification.
> +        \item If the number of descriptors added to the queue since
> +          the last notification is greater than 65535, the driver MUST
> +          send a notification.
>          \item Otherwise the driver SHOULD NOT send a notification.
>    \end{itemize}
>  \end{itemize}
> -- 
> 2.32.0 (Apple Git-131)
> 
> 
> -Patrick Mosca

I am not sure I understand. Can you give an example of how the issue
triggers please?

-- 
MST