Re: [PATCH] xenvif: fix TX doorbell coalescing for ALGORITHM_NONE traffic

[Mounah Honore Cesaire <cesaire.mounah@xxxxxxxxxx>]

On 9/25/26 4:02 PM, david ambu wrote:
> Each NBL carries a single hash value covering all its NET_BUFFERs.
> All NB within the same NBL therefore map to the same TX ring, so the
> ring doorbell can be deferred across NB and fired once at the NBL
> boundary when More=FALSE.
>
> TransmitterQueuePacket was overriding More=FALSE unconditionally in
> the XENVIF_PACKET_HASH_ALGORITHM_NONE case, causing a ring kick on
> every individual NET_BUFFER instead of once per NBL.
>
> Removing the override lets the More flag from xennet flow through
> correctly, so ALGORITHM_NONE traffic gets the same per-NBL kick
> behaviour as ALGORITHM_TOEPLITZ.
>
> Signed-off-by: david ambu <david.preetham@xxxxxxxxxx>
> Co-Authored-By: Claude Sonnet 4.6 <noreply@xxxxxxxxxxxxx>
> ---
>   src/xenvif/transmitter.c | 1 -
>   1 file changed, 1 deletion(-)
>
> diff --git a/src/xenvif/transmitter.c b/src/xenvif/transmitter.c
> index 8c4ee41..ad58336 100644
> --- a/src/xenvif/transmitter.c
> +++ b/src/xenvif/transmitter.c
> @@ -5245,7 +5245,6 @@ TransmitterQueuePacket(
>       switch (Algorithm) {
>       case XENVIF_PACKET_HASH_ALGORITHM_NONE:
>           Value = __TransmitterHashPacket(Transmitter, Packet);
> -        More = FALSE;
>           break;
>   
>       case XENVIF_PACKET_HASH_ALGORITHM_UNSPECIFIED:

Reviewed-by: Césaire Mounah <cesaire.mounah@xxxxxxxxxx>

-- 
Césaire Mounah | Vates Windows Guest Tools Engineer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech



--
Césaire Mounah | Vates Windows Guest Tools Engineer

XCP-ng & Xen Orchestra - Vates solutions

web: https://vates.tech