Re: [Xen-devel] [PATCH] libxc: migrate migration stream definitions into Xen public headers

[David Woodhouse <dwmw2@xxxxxxxxxxxxx>]

On 1 February 2020 11:45:42 GMT, "Durrant, Paul" <pdurrant@xxxxxxxxxxxx> wrote:
>> -----Original Message-----
>> From: David Woodhouse <dwmw2@xxxxxxxxxxxxx>
>> Sent: 01 February 2020 00:40
>> To: xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
>> Cc: Ian Jackson <ian.jackson@xxxxxxxxxxxxx>; Wei Liu <wl@xxxxxxx>;
>Andrew
>> Cooper <andrew.cooper3@xxxxxxxxxx>; George Dunlap
>> <George.Dunlap@xxxxxxxxxxxxx>; Jan Beulich <jbeulich@xxxxxxxx>;
>Julien
>> Grall <julien@xxxxxxx>; Konrad Rzeszutek Wilk
><konrad.wilk@xxxxxxxxxx>;
>> Stefano Stabellini <sstabellini@xxxxxxxxxx>; Gautam, Varad
>> <vrd@xxxxxxxxx>; Durrant, Paul <pdurrant@xxxxxxxxxxxx>
>> Subject: [PATCH] libxc: migrate migration stream definitions into Xen
>> public headers
>> 
>> From: David Woodhouse <dwmw@xxxxxxxxxxxx>
>> 
>> These data structures will be used for live update, which is
>basically
>> just live migration from one Xen to the next on the same machine via
>> in-memory data structures, instead of across the network.
>> 
>> Signed-off-by: David Woodhouse <dwmw@xxxxxxxxxxxx>
>> Well-excellent-carry-on-then-by: Ian Jackson
><ian.jackson@xxxxxxxxxxxxx>
>> Go-with-that-for-now-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>
>This is probably fine but may need some re-structuring when we consider
>non-cooperative migration. I have not yet had time to post a design
>document but my thoughts are along the lines of generalising the idea
>of 'hvm_context' records into 'domain_context' (or some suitable name).
>Then we can add the necessary 'PV' save records, such as event channel
>state, grant table state, etc. We may then want to consider using the
>same mechanism for PV shared info and maybe some of the other records
>that are currently 'top level' in the XC migration stream.

Right. This is just the existing ABI; structures that exist on the wire and are 
in some cases exported by existing domctls.

As we build upon it for live update and GT migration, we need to carefully 
consider how we do that — and whether any of the new or indeed existing fields 
in the HVM_CONTEXT should actually be top-level migration records in their own 
right, which I think is what you're suggesting.

This patch is just the basis for that work, as it's precisely the structures we 
have already; just renamed and moved into scope for what comes next.

-- 
dwmw2

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel