On Mon, Feb 27, 2023 at 03:02:10AM +0000, Parav Pandit wrote:
>
> > From: Michael S. Tsirkin <[email protected]>
> > Sent: Saturday, February 25, 2023 6:09 PM
> > On Sun, Feb 26, 2023 at 12:29:59AM +0200, Parav Pandit wrote:
> > > Legacy interface PCI Device layout description has following issues.
> > >
> > > 1. repeated 'structure' word
> > > 2. virtio header was defined the 0.9.5 spec. In a legacy interface
> > > section it is referred with different keywards
> > > as (a) virtio header, (b) general headers, (c) legacy configuration
> > > structure, (d) virtio common configuration structure and
> > > (e) other fields.
> > > 3. Driver and device requirements listing is intermixed.
> > > 4. spelling error of structure
> > >
> > > Hence, rewrite the description to eliminate above issues as below.
> > >
> > > 1. Removed repeated structure word
> > > 2. Fix spelling of structure
> > > 3. Place all device requirements toegether 3. Define legacy
> > > configuration structure that consist of
> > > a. legacy common configuration structure and
> > > b. device specific configuration structure 4. Rewrite section
> > > around above changes
> > >
> > > This is only an editorial change.
> >
> > No, editorial changes are things like spelling corrections.
> >
> Got it. Will drop this remark.
>
> > I have trouble reviewing
> > because I have no idea why you are making each change.
> >
> After documenting legacy interface in the spec, we cannot claim that driver
> is the source = specification.
We document the *legacy interface*. What we don't is we don't document
legacy devices and drivers:
Legacy devices and legacy drivers are not compliant with this
specification.
> Currently transitional device and pci device do not scale well (or doesn't
> scale at all).
> We are working on supporting it and having better defined transitional device
> seems important part of it.
>
> > E.g. you just rewrote a bunch of text and I frankly don't know why. For
> > example:
> >
> > -When using the legacy interface, transitional drivers
> > -MUST use the legacy configuration structure in BAR0 in the first
> > -I/O region of the PCI device, as documented below.
> >
> > is now apparently:
> >
> > +When used through the legacy interface, the legacy common
> > +configuration structure has the following layout:
> >
> > and this is better? why? we are replacing a clear requirement which applied
> > to
>
> Because as I explained in the commit log, that one structure is named using 5
> different names.
So maybe try just renaming. What you did is also convert a normative
statement to a non-normative one.
> > drivers with a vague statement which I can't say what it applies to.
> >
> > Any chance of splitting these things up? Maybe that will help.
> >
> That's a good idea. Yes. I will split these patches to smaller one.
>
> > Apropos I don't know that we want to spend much time on legacy sections.
> Transitional devices are very much in use and documenting them well is
> important to build scalable transitional devices.
>
> > Really with legacy code is the main source of documentation - if drivers
> > work
> > then you are golden. If they don't appealing to spec will not help.
>
> Yes, so don't want to add additional text. Just correcting the current one.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
