[ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

Folks,

With 64-bit support well established in the x86 world these days, the number
of x86 production environments that cannot run a 64-bit hypervisor is pretty
much nil. Maintaining the 32-bit x86 port, and implementing new features for
it, is an ongoing development burden which could be more usefully directed
elsewhere. Therefore, 32-bit x86 will be considered obsolete in the 4.3
development branch, and removed.

32-bit x86 will continue to be maintained and supported (insofar as it has
been recently) in the supported stable branches: 3.4, 4.1, and 4.2.

Furthermore, 32-bit guests (including 32-bit PV guests) will continue to be
supported on the 64-bit hypervisor, as they always have been.

 Regards,
 Keir & The Xen Team

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 09/09/2012 12:11, "Keir Fraser" <keir.xen@gmail.com> wrote:

> Folks,
> 
> With 64-bit support well established in the x86 world these days, the number
> of x86 production environments that cannot run a 64-bit hypervisor is pretty
> much nil. Maintaining the 32-bit x86 port, and implementing new features for
> it, is an ongoing development burden which could be more usefully directed
> elsewhere. Therefore, 32-bit x86 will be considered obsolete in the 4.3
> development branch, and removed.

Ian,

Please can you remove x86_32 from the automated tests for xen-unstable?

 Thanks,
 Keir

> 32-bit x86 will continue to be maintained and supported (insofar as it has
> been recently) in the supported stable branches: 3.4, 4.1, and 4.2.
> 
> Furthermore, 32-bit guests (including 32-bit PV guests) will continue to be
> supported on the 64-bit hypervisor, as they always have been.
> 
>  Regards,
>  Keir & The Xen Team
> 
> 

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Ian Jackson]

Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2"):
> Please can you remove x86_32 from the automated tests for xen-unstable?

I will do this but I have a number of other things which need to go
through the autotester's tests of itself, first.  I'll let you know
when it's done.

Thanks,
Ian.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 10/09/2012 15:16, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:

> Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life
> after 4.2"):
>> Please can you remove x86_32 from the automated tests for xen-unstable?
> 
> I will do this but I have a number of other things which need to go
> through the autotester's tests of itself, first.  I'll let you know
> when it's done.

A self-aware autotester. :)

Well that's fine. It doesn't have to be done in a huge hurry.

 Thanks,
 Keir

> Thanks,
> Ian.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[David Vrabel]

On 09/09/12 12:11, Keir Fraser wrote:
> Folks,
> 
> With 64-bit support well established in the x86 world these days, the number
> of x86 production environments that cannot run a 64-bit hypervisor is pretty
> much nil. Maintaining the 32-bit x86 port, and implementing new features for
> it, is an ongoing development burden which could be more usefully directed
> elsewhere. Therefore, 32-bit x86 will be considered obsolete in the 4.3
> development branch, and removed.

I have some tracing patches that have some 32-bit/64-bit #ifdef'ery.
Would you prefer them posted now, or deferred until 32-bit is gone?

David

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 10/09/2012 17:24, "David Vrabel" <david.vrabel@citrix.com> wrote:

> On 09/09/12 12:11, Keir Fraser wrote:
>> Folks,
>> 
>> With 64-bit support well established in the x86 world these days, the number
>> of x86 production environments that cannot run a 64-bit hypervisor is pretty
>> much nil. Maintaining the 32-bit x86 port, and implementing new features for
>> it, is an ongoing development burden which could be more usefully directed
>> elsewhere. Therefore, 32-bit x86 will be considered obsolete in the 4.3
>> development branch, and removed.
> 
> I have some tracing patches that have some 32-bit/64-bit #ifdef'ery.
> Would you prefer them posted now, or deferred until 32-bit is gone?

If it's not too intricate then post them now, and we'll strip out the 32-bit
parts later. I don't think we should hold up pending patches for this.
Although I'm holding back some further development in anticipation of it. :)

 -- Keir

> David

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Ian Jackson]

Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2"):
> On 10/09/2012 15:16, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:
> > I will do this but I have a number of other things which need to go
> > through the autotester's tests of itself, first.  I'll let you know
> > when it's done.
> 
> A self-aware autotester. :)

As you say.

> Well that's fine. It doesn't have to be done in a huge hurry.

I've been looking into this and have a draft of a suitable change.
But: can you confirm that typing "make" will still work on a 32-bit
system and generate a working tools build ?  I don't care what
hypervisor (if any) it builds, but it has to succeed because that's
how I build the tools for use in 32-bit dom0s.

Ian.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 10/09/2012 18:09, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:

> Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life
> after 4.2"):
>> On 10/09/2012 15:16, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:
>>> I will do this but I have a number of other things which need to go
>>> through the autotester's tests of itself, first.  I'll let you know
>>> when it's done.
>> 
>> A self-aware autotester. :)
> 
> As you say.
> 
>> Well that's fine. It doesn't have to be done in a huge hurry.
> 
> I've been looking into this and have a draft of a suitable change.
> But: can you confirm that typing "make" will still work on a 32-bit
> system and generate a working tools build ?  I don't care what
> hypervisor (if any) it builds, but it has to succeed because that's
> how I build the tools for use in 32-bit dom0s.

You want 'make' at the root Makefile to build successfully? If so we can
stub out 32-bit hypervisor only either in that Makefile, or xen/Makefile.

> Ian.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Ian Jackson]

Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2"):
> On 10/09/2012 18:09, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:
> > I've been looking into this and have a draft of a suitable change.
> > But: can you confirm that typing "make" will still work on a 32-bit
> > system and generate a working tools build ?  I don't care what
> > hypervisor (if any) it builds, but it has to succeed because that's
> > how I build the tools for use in 32-bit dom0s.
> 
> You want 'make' at the root Makefile to build successfully? If so we can
> stub out 32-bit hypervisor only either in that Makefile, or xen/Makefile.

That would be best for me.  I think it would be best for users too.
Maybe you would like to leave a xen-blah.README in dist/install/boot.

Ian.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 11/09/2012 11:27, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:

> Keir Fraser writes ("Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life
> after 4.2"):
>> On 10/09/2012 18:09, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:
>>> I've been looking into this and have a draft of a suitable change.
>>> But: can you confirm that typing "make" will still work on a 32-bit
>>> system and generate a working tools build ?  I don't care what
>>> hypervisor (if any) it builds, but it has to succeed because that's
>>> how I build the tools for use in 32-bit dom0s.
>> 
>> You want 'make' at the root Makefile to build successfully? If so we can
>> stub out 32-bit hypervisor only either in that Makefile, or xen/Makefile.
> 
> That would be best for me.  I think it would be best for users too.
> Maybe you would like to leave a xen-blah.README in dist/install/boot.

Okay, I'll stub out in xen/Makefile and install something into
dist/install/boot.

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Jan Beulich]

>>> On 11.09.12 at 12:44, Keir Fraser <keir@xen.org> wrote:
> Okay, I'll stub out in xen/Makefile and install something into
> dist/install/boot.

If you're already in the process of doing this, may I ask that you
please don't remove xen/arch/x86/cpu/centaur.c (as a pending
patch of mine will make use of this for properly enabling 64-bit
support on VIA CPUs)?

Jan

Re: [ANNOUNCE] Xen x86 32-bit hypervisor end of life after 4.2

[Keir Fraser]

On 11/09/2012 11:51, "Jan Beulich" <JBeulich@suse.com> wrote:

>>>> On 11.09.12 at 12:44, Keir Fraser <keir@xen.org> wrote:
>> Okay, I'll stub out in xen/Makefile and install something into
>> dist/install/boot.
> 
> If you're already in the process of doing this, may I ask that you
> please don't remove xen/arch/x86/cpu/centaur.c (as a pending
> patch of mine will make use of this for properly enabling 64-bit
> support on VIA CPUs)?

I'm waiting for the okay from IanJ before I start. And, okay, I won't remove
centaur.c. I will probably just do a rough prune to start with anyway, then
others can go in and help finish the job off.

 -- Keir

> Jan
>