Re: [virtio-comment] [PATCH 1/1] RFC: virtio-bt: add virtio BT device specification

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

On Wed, Mar 15, 2023 at 04:55:59PM +0100, Cornelia Huck wrote:
> On Fri, Mar 10 2023, Igor Skalkin <Igor.Skalkin@opensynergy.com> wrote:
> 
> > This PR is aimed as review for comments(RFC) purpose.
> >
> > * Initial draft version.
> >
> > Signed-off-by: Igor Skalkin <Igor.Skalkin@opensynergy.com>
> > ---
> >  conformance.tex                        |  12 ++-
> >  content.tex                            |   1 +
> >  device-types/bt/description.tex        | 112 +++++++++++++++++++++++++
> >  device-types/bt/device-conformance.tex |   8 ++
> >  device-types/bt/driver-conformance.tex |   8 ++
> >  5 files changed, 137 insertions(+), 4 deletions(-)
> >  create mode 100644 device-types/bt/description.tex
> >  create mode 100644 device-types/bt/device-conformance.tex
> >  create mode 100644 device-types/bt/driver-conformance.tex
> 
> (...)
> 
> > diff --git a/device-types/bt/description.tex b/device-types/bt/description.tex
> > new file mode 100644
> > index 0000000..3ce265d
> > --- /dev/null
> > +++ b/device-types/bt/description.tex
> > @@ -0,0 +1,112 @@
> > +\section{BT Device}\label{sec:Device Types / BT Device}
> > +
> > +The virtio-bt device provides an HCI (Host Control Interface) over VirtIO
> > +link between the guest HCI device and the host HCI backend.
> > +Also, the device can inform the guest driver which vendor-specific command
> > +set it supports.
> > +Host Control Interface is described in Bluetooth Core Specification:
> > +\newline\url{https://www.bluetooth.com/specifications/specs/core-specification-5-4/}\\
> 
> I guess that document describes what the device/driver MUST implement?
> If so, I think it needs to be added to the "Normative References"
> section in introduction.tex.
> 
> > +
> > +\subsection{Device ID}\label{sec:Device Types / BT Device / Device ID}
> > +
> > +40
> > +
> > +\subsection{Virtqueues}\label{sec:Device Types / BT Device / Virtqueues}
> > +
> > +\begin{description}
> > +\item[0] transmitq
> > +\item[1] receiveq
> > +\end{description}
> > +
> > +\subsection{Feature bits}\label{sec:Device Types / BT Device / Feature bits}
> > +
> > +\begin{description}
> > +\item[VIRTIO_BT_F_VND_HCI (0)]  Indicates vendor command support.
> > +\item[VIRTIO_BT_F_MSFT_EXT (1)] Indicates MSFT vendor support.
> > +\item[VIRTIO_BT_F_AOSP_EXT (2)] Indicates AOSP vendor support.
> > +\item[VIRTIO_BT_F_CONFIG_V2 (3)] The device uses the second version of the
> > +configuration space structure.
> > +\end{description}
> > +
> > +\devicenormative{\subsubsection}{Feature bits}{Device Types / BT Device / Feature bits}
> > +
> > +The device MUST require the driver to accept the VIRTIO_BT_F_CONFIG_V2 feature
> > +bit, i.e. not set FEATURES_OK without it, and use the second version
> > +(struct virtio_bt_config_v2) of the configuration layout, because the
> > +first one (struct virtio_bt_config) is unaligned, which violates the
> > +specification.
> 
> Did we have a device or driver that didn't use v2? I'm not sure we want
> to add a feature for that, other than for backwards compatibility.

Linux drivers use a different layout, yes.

I think it should be possible to implement device without
VIRTIO_BT_F_CONFIG_V2 if someone wants to be compatible.

And hmm we need to get back to addressing the negotiation mess ...



> > +
> > +The device should offer VIRTIO_BT_F_MSFT_EXT or VIRTIO_BT_F_AOSP_EXT feature bit
> > +if it supports correspondingly MSFT or AOSP extension command sets. In case of
> > +VIRTIO_BT_F_MSFT_EXT, device should also set configuration \field{msft_opcode}.
> > +
> > +The device should offer VIRTIO_BT_F_VND_HCI feature bit and set \field{vendor}
> > +to the VIRTIO_BT_CONFIG_VENDOR_ZEPHYR, VIRTIO_BT_CONFIG_VENDOR_INTEL or
> > +VIRTIO_BT_CONFIG_VENDOR_REALTEK, if it supports corresponding vendor extensions.
> 
> Where are those extension command sets and vendor extensions
> described - in the Core Specifications linked above?
> 
> > +
> > +\drivernormative{\subsubsection}{Feature bits}{Device Types / BT Device / Feature bits}
> > +
> > +The driver MUST accept VIRTIO_BT_F_CONFIG_V2 feature bit if offered by the device.
> > +
> > +The driver SHOULD accept any of the  VIRTIO_BT_F_VND_HCI, VIRTIO_BT_F_MSFT_EXT
> > +or VIRTIO_BT_F_AOSP_EXT feature bits if offered by the device.
> > +
> > +\subsection{Device configuration layout}\label{sec:Device Types / BT Device / Device configuration layout}
> > +
> > +
> > +The first version:
> > +\begin{lstlisting}
> > +struct virtio_bt_config {
> > +	u8 type;
> > +	le16 vendor;
> > +	le16 msft_opcode;
> > +} __attribute__((packed));
> > +\end{lstlisting}
> > +
> > +is deprecated, new devices must use the second one:
> > +\begin{lstlisting}
> > +struct virtio_bt_config_v2 {
> > +	u8 type;
> > +	u8 alignment;
> > +	le16 vendor;
> > +	le16 msft_opcode;
> > +};
> > +\end{lstlisting}
> > +
> > +\devicenormative{\subsubsection}{Device configuration layout}{Device Types / BT Device / Device configuration layout}
> > +The device MUST NOT present a value different than
> > +\begin{lstlisting}
> > +	VIRTIO_BT_CONFIG_TYPE_PRIMARY = 0,
> > +	VIRTIO_BT_CONFIG_TYPE_AMP = 1,
> > +\end{lstlisting}
> > +in \field{type}.
> 
> I think it would be better to move this out of the normative section and
> use something like
> 
> "
> The \field{type} field can have the following values:
> 
> \begin{lstlisting}
> #define VIRTIO_BT_CONFIG_TYPE_PRIMARY 0
> #define VIRTIO_BT_CONFIG_TYPE_AMP     1
> \end{lstlisting}
> "
> 
> I don't think we need to bother with stating explicitly that the device
> MUST NOT use any undefined values.
> 
> > +
> > +The values 1..3 of the \field{vendor} are already reserved for vendor extensions listed below:


don't repeat 1..3 here - either use latex trickery or just omit
as we'll otherwise forget to update this.

> > +\begin{lstlisting}
> > +	VIRTIO_BT_CONFIG_VENDOR_NONE = 0
> > +	VIRTIO_BT_CONFIG_VENDOR_ZEPHYR = 1
> > +	VIRTIO_BT_CONFIG_VENDOR_INTEL = 2
> > +	VIRTIO_BT_CONFIG_VENDOR_REALTEK = 3
> > +\end{lstlisting}
> 
> Same here.
> 
> I guess the various vendor extensions are mutually exclusive?

are the extensions in the specification linked?

> > +
> > +If value of the \field{vendor} is not VIRTIO_BT_CONFIG_VENDOR_NONE, device MUST
> > +offer VIRTIO_BT_F_VND_HCI feature bit.
> 
> Maybe
> 
> "The device MUST offer the VIRTIO_BT_F_VND_HCI feature bit if it sets
> \field{vendor} to any value other than VIRTIO_BT_CONFIG_VENDOR_NONE."
> 
> ?
> 
> > +
> > +\drivernormative{\subsubsection}{Driver configuration layout}{Device Types / BT Device / Driver configuration layout}
> > +All configuration fields are read-only for the driver.
> 
> This isn't a normative statement -- move it to the non-normative
> section?
> 
> > +
> > +\subsection{Device initialization}\label{sec:Device Types / BT Device / Device initialization}
> > +
> > +The virtqueues are initialized.
> > +
> > +\subsection{Device operations}\label{sec:Device Types / BT Device / Device operations}
> > +
> > +The driver SHOULD populate the receive queue with at least one buffer of at
> 
> "The driver populates" ?
> 
> > +least 258 bytes to contain 1 byte "packet type" and HCI event packet (2 bytes
> > +of HCI event packet header and up to 255 bytes payload).
> > +Synchronous and asynchronous data packets that are longer than the provided
> > +buffer will be fragmented.
> > +
> > +The driver sends to the transmit queue all (command and data) packets, received
> > +from the guest HCI device, and transfers to the guest HCI device all (event and
> > +data) HCI packets, received from the receive queue.