PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Keir Fraser]

Folks,

There's been discussion in the past about eventually removing support for
32-bit x86 in the development branch of the hypervisor. The reasons being:
 * It takes developer effort to maintain and test
 * Most repackagers are supplying 64-bit only hypervisor binary these days
 * 32-bit target is not supported in any case for various new hypervisor
features (e.g., memory paging, memory sharing, memory access trapping)
 * 64-bit support has been standard in all mainstream x86 processors for
about the last 5 years. With a 64-bit processor there is no reason to run
the 32-bit hypervisor, as the 64-bit hypervisor supports all 32-bit guests.

With 4.1 branched I'd like to raise this topic again. Does anyone have a
strong argument for keeping the 32-bit x86 target alive?

 Thanks,
 Keir

Re: PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Ian Jackson]

Keir Fraser writes ("[Xen-devel] PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch"):
> With 4.1 branched I'd like to raise this topic again. Does anyone have a
> strong argument for keeping the 32-bit x86 target alive?

"Mini" laptops without 64-bit support are available for sale from many
vendors including Toshiba, Asus, Acer, etc.

Ian.

Re: PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Keir Fraser]

On 17/03/2011 17:11, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:

> Keir Fraser writes ("[Xen-devel] PROPOSAL: Remove 32-bit x86 host support in
> 4.2 development branch"):
>> With 4.1 branched I'd like to raise this topic again. Does anyone have a
>> strong argument for keeping the 32-bit x86 target alive?
> 
> "Mini" laptops without 64-bit support are available for sale from many
> vendors including Toshiba, Asus, Acer, etc.

Yes, we might end up keeping x32 support, even if only as a poor cousin, for
this reason. We don't really know what the EOL is for 32-bit-only x86
processors and someone may care about targeting these low-end netbooks and
tablets at some point. Much easier to spruce up existing flagging x32
support than re-fit it from scratch.

 -- Keir

> Ian.

Re: PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Ian Jackson]

Keir Fraser writes ("Re: [Xen-devel] PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch"):
> Yes, we might end up keeping x32 support, even if only as a poor cousin, for
> this reason. We don't really know what the EOL is for 32-bit-only x86
> processors and someone may care about targeting these low-end netbooks and
> tablets at some point. Much easier to spruce up existing flagging x32
> support than re-fit it from scratch.

I think that would be wise.

Ian.

Re: PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Christian Tramnitz]

On 17.03.2011 18:11 Ian Jackson wrote:
> "Mini" laptops without 64-bit support are available for sale from many
> vendors including Toshiba, Asus, Acer, etc.

This can only be first gen (N2xx series) Atoms, which have been 
superseded by N4xx/N5xx (supporting x64) in 2009 or Atom Z-series 
(targeted at MIDs).
Those first gen N2xx have no VT (or PM) support either, so I do not see 
this as real target platform for upcoming Xen major releases where new 
features wouldn't be available for this limited CPU feature set anyway...
(I'm not saying you shouldn't run Xen on a Netbook, but what would be 
the advantage of upgrading from Xen 4.1.x to 4.2.x on a 
x86_32/no-VT/no-PM device?)

Best regards,
    Christian

Re: [Xen-devel] Re: PROPOSAL: Remove 32-bit x86 host support in 4.2 development branch

[Alan Cox]

> superseded by N4xx/N5xx (supporting x64) in 2009 or Atom Z-series 
> (targeted at MIDs).

And found in systems like the original Dell mini-10, which I've never
tried with Xen but does handle kvm quite happily.