RE: [virtio-dev] Re: [PATCH v1 1/2] transport-pci: Introduce legacy registers access commands

[Parav Pandit <parav@nvidia.com>]

> From: Michael S. Tsirkin <mst@redhat.com>
> Sent: Wednesday, May 3, 2023 4:09 PM
> 
> On Wed, May 03, 2023 at 03:38:31PM -0400, Parav Pandit wrote:
> >
> >
> > On 5/3/2023 3:21 PM, Michael S. Tsirkin wrote:
> > > On Wed, May 03, 2023 at 06:26:58AM +0300, Parav Pandit wrote:
> > > > +\begin{lstlisting}
> > > > +struct virtio_admin_cmd_lreg_wr_data {
> > > > +	u8 offset; /* Starting byte offset of the register(s) to write */
> > > > +	u8 size; /* Number of bytes to write into the register. */
> > > > +	u8 register[];
> > > > +};
> > >
> > > BTW if we don't have padding we could reuse buffer size and won't
> > > need size here.
> > Yes. sounds good.
> >
> > > Do we want to tweak admin command structure generally to use
> > > u8 and not le64 for command data then? WDYT?
> > >
> > yeah, cmd and result be u8 is fine.
> > If padding is needed for perf, may be that command can define the padding.
> 
> The commands that I defined do need u64 (much easier to work on bitmaps in
> fixed size chunks), so they will use that.

You have already defined it as le64 field. See it below.
So it is not padding, the field itself is le64.
So no need to typecast it differently based on the descriptor size because it is always array of le64.
It behaves as padding when number of bits are < 64, but it is really not a (dynamic) padding.

le64 device_admin_cmd_opcodes[];