Re: Using param structure in VIF interface

["Durrant, Paul" <xadimgnik@xxxxxxxxx>]

On 14/06/2022 09:55, Martin Harvey wrote:
> -----Original Message-----
> From: Durrant, Paul <xadimgnik@xxxxxxxxx>
> Sent: 13 June 2022 17:34
>
> > Just to let you know that I am working on this...
> Okay! <Relived!>
>
> > It's more of a can of worms than I thought... Now that I have split up the 
> > patches I realise that, to use a params structure, we need that to be versioned 
> > along with the interface.
> All right. I reckoned we weren't going to change it that much, and that we 
> could possibly cover all the changes under the VIF interface version number. I 
> know individual versioning of structures is a thing, but we need not slavlishly 
> apply it to everything.
Versioning needs to be well-structured though, and in this case the 
cascading of the callback variants allows me to put the XENVIF-only 
back-pressure code in the 'right' place i.e. in the compatibility shim.
> > Also when XENNET is using a down-rev version, which could theoretically be 
> > behind the latest version in XENVIF, we are going to need a chain of callback 
> > functions to translate each version of the params structure to/from the 
> > previous version until we get to the one that XENNET actually deals with.
> Yes, that's a possible use case, but hopefully we would know what versions use 
> what. Again, I was hoping that we weren't going to change the callback params 5 
> or 6 times in quick succession.
>
> > I have something partially done, but it's going to take another day or so.
> OK, thank you.
Patches posted... 4 in XENVIF (including an unrelated ASSERTion fix and 
retirement patches) and 2 in XENNET for the two interface version bumps 
(so I could check that different interface combos worked correctly).
  Paul