[PATCH v2] virtio-net: Ignore num_buffers when unused

[PATCH v2] virtio-net: Ignore num_buffers when unused

[Akihiko Odaki]

If VIRTIO_NET_F_MRG_RXBUF has not been negotuated, vhost_net and QEMU
leaves num_buffers uninitialized while the specification tells to set it
to one.

Let drivers ignore the field to make them compatible with
vhost_net/QEMU, in a way similar to hdr_len field, which is also
marked as unreliable due to implementation bugs.

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
---
Changes in v2:
- Corrected the change description to tell vhost_net and QEMU leaves
  num_buffers uninitialized where it used to tell they set it to zero.
- Restored statements that require devices to set num_buffers to one.
- Link to v1: https://lore.kernel.org/r/20241226-reserved-v1-1-39ca566889c0@daynix.com
---
 device-types/net/description.tex | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/device-types/net/description.tex b/device-types/net/description.tex
index 76585b0..70c8b31 100644
--- a/device-types/net/description.tex
+++ b/device-types/net/description.tex
@@ -863,6 +863,12 @@ \subsubsection{Processing of Incoming Packets}\label{sec:Device Types / Network
 This is due to various bugs in implementations.
 \end{note}
 
+If VIRTIO_NET_F_MRG_RXBUF has not been negotiated, the driver MUST ignore
+\field{num_buffers}.
+\begin{note}
+This is due to bugs in implementations.
+\end{note}
+
 If neither VIRTIO_NET_HDR_F_NEEDS_CSUM nor
 VIRTIO_NET_HDR_F_DATA_VALID is set, the driver MUST NOT
 rely on the packet checksum being correct.

---
base-commit: b495841a8e80d12c1130f8868f4128866291142d
change-id: 20241226-reserved-d5960f173223

Best regards,
-- 
Akihiko Odaki <akihiko.odaki@daynix.com>

Re: [PATCH v2] virtio-net: Ignore num_buffers when unused

[Michael S. Tsirkin]

On Fri, Dec 27, 2024 at 01:08:50PM +0900, Akihiko Odaki wrote:
> If VIRTIO_NET_F_MRG_RXBUF has not been negotuated, vhost_net and QEMU
> leaves num_buffers uninitialized while the specification tells to set it
> to one.
> 
> Let drivers ignore the field to make them compatible with
> vhost_net/QEMU, in a way similar to hdr_len field, which is also
> marked as unreliable due to implementation bugs.
> 
> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> ---
> Changes in v2:
> - Corrected the change description to tell vhost_net and QEMU leaves
>   num_buffers uninitialized where it used to tell they set it to zero.
> - Restored statements that require devices to set num_buffers to one.
> - Link to v1: https://lore.kernel.org/r/20241226-reserved-v1-1-39ca566889c0@daynix.com
> ---
>  device-types/net/description.tex | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/device-types/net/description.tex b/device-types/net/description.tex
> index 76585b0..70c8b31 100644
> --- a/device-types/net/description.tex
> +++ b/device-types/net/description.tex
> @@ -863,6 +863,12 @@ \subsubsection{Processing of Incoming Packets}\label{sec:Device Types / Network
>  This is due to various bugs in implementations.
>  \end{note}
>  
> +If VIRTIO_NET_F_MRG_RXBUF has not been negotiated, the driver MUST ignore
> +\field{num_buffers}.
> +\begin{note}
> +This is due to bugs in implementations.
> +\end{note}
> +
>  If neither VIRTIO_NET_HDR_F_NEEDS_CSUM nor
>  VIRTIO_NET_HDR_F_DATA_VALID is set, the driver MUST NOT
>  rely on the packet checksum being correct.


should

we can not post factum make existing previously compliant drivers non-complient

> ---
> base-commit: b495841a8e80d12c1130f8868f4128866291142d
> change-id: 20241226-reserved-d5960f173223
> 
> Best regards,
> -- 
> Akihiko Odaki <akihiko.odaki@daynix.com>