2.6.20.3 to be the next kernel?

2.6.20.3 to be the next kernel?

[Tim Post]

Hello to all :

I found this commitment from a development fedora repo :

http://www.redhat.com/archives/fedora-extras-commits/2007-May/msg02474.html

On may 10, looks like fedora is about to go to 2.6.20.3 for its xen
kernel (sometime in the not too distant future).

I looked through the xen-devel archives for mentions of 2.6.20.3, next
xen kernel, next kernel, to see if this was a popular question. I did
not see much mention of anything, so I apologize if I'm asking something
often answered. I only found mentions of 2.6.20 from a couple months
back.

Where is the next (planned) hop in Linux kernels going to lead?

I'm not a list subscriber, please cc me if replying.

Thanks in advance, 
--Tim

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

> I found this commitment from a development fedora repo :
>
> http://www.redhat.com/archives/fedora-extras-commits/2007-May/msg02474.html
>
> On may 10, looks like fedora is about to go to 2.6.20.3 for its xen
> kernel (sometime in the not too distant future).

I believe the Fedora folks forward-port the Xen patches to newer kernels, so 
this doesn't necessarily indicate that the mainline Xen tree will be moving 
to this kernel version.

> Where is the next (planned) hop in Linux kernels going to lead?

Not sure.  A few interesting things are currently in progress with respect to 
kernel versions; the Xen paravirt-ops patches may be going into the next 
release mainline Linux (maybe?  I think?) at which point you'll be able to 
build domU kernels directly from kernel.org sources.

Secondly we're going to move away from the sparse tree layout (yay) and 
towards having a separate repository for the XenLinux tree.

Cheers,
Mark

-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

Re: 2.6.20.3 to be the next kernel?

[Daniel P. Berrange]

On Mon, Jun 04, 2007 at 07:01:22PM +0100, Mark Williamson wrote:
> > I found this commitment from a development fedora repo :
> >
> > http://www.redhat.com/archives/fedora-extras-commits/2007-May/msg02474.html
> >
> > On may 10, looks like fedora is about to go to 2.6.20.3 for its xen
> > kernel (sometime in the not too distant future).
> 
> I believe the Fedora folks forward-port the Xen patches to newer kernels, so 
> this doesn't necessarily indicate that the mainline Xen tree will be moving 
> to this kernel version.

Indeed - this is a forward port of the mainline Xen 2.6.18 to newer LKML
base, its not a vanilla Xen tree.

> > Where is the next (planned) hop in Linux kernels going to lead?
> 
> Not sure.  A few interesting things are currently in progress with respect to 
> kernel versions; the Xen paravirt-ops patches may be going into the next 
> release mainline Linux (maybe?  I think?) at which point you'll be able to 
> build domU kernels directly from kernel.org sources.
> 
> Secondly we're going to move away from the sparse tree layout (yay) and 
> towards having a separate repository for the XenLinux tree.

  s/are going to move/have moved/   :-)

changeset 15204: 	1712c62b913c
         author: 	Ian Campbell <ian.campbell@xensource.com>
           date: 	Mon Jun 04 11:16:19 2007 +0100


Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

Re: 2.6.20.3 to be the next kernel?

[Nate Carlson]

On Mon, 4 Jun 2007, Mark Williamson wrote:
> Not sure.  A few interesting things are currently in progress with 
> respect to kernel versions; the Xen paravirt-ops patches may be going 
> into the next release mainline Linux (maybe?  I think?) at which point 
> you'll be able to build domU kernels directly from kernel.org sources.

Just curious - can you also build dom0?  :)

------------------------------------------------------------------------
| nate carlson | natecars@natecarlson.com | http://www.natecarlson.com |
|       depriving some poor village of its idiot since 1981            |
------------------------------------------------------------------------

Re: 2.6.20.3 to be the next kernel?

[Daniel P. Berrange]

On Mon, Jun 04, 2007 at 01:13:01PM -0500, Nate Carlson wrote:
> On Mon, 4 Jun 2007, Mark Williamson wrote:
> >Not sure.  A few interesting things are currently in progress with 
> >respect to kernel versions; the Xen paravirt-ops patches may be going 
> >into the next release mainline Linux (maybe?  I think?) at which point 
> >you'll be able to build domU kernels directly from kernel.org sources.
> 
> Just curious - can you also build dom0?  :)

No. AFAIK, at this point in time, Xen paravirt_ops is DomU only. 

Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

> > > Where is the next (planned) hop in Linux kernels going to lead?
> >
> > Not sure.  A few interesting things are currently in progress with
> > respect to kernel versions; the Xen paravirt-ops patches may be going
> > into the next release mainline Linux (maybe?  I think?) at which point
> > you'll be able to build domU kernels directly from kernel.org sources.
> >
> > Secondly we're going to move away from the sparse tree layout (yay) and
> > towards having a separate repository for the XenLinux tree.
>
>   s/are going to move/have moved/   :-)
>
> changeset 15204: 	1712c62b913c
>          author: 	Ian Campbell <ian.campbell@xensource.com>
>            date: 	Mon Jun 04 11:16:19 2007 +0100


AWESOME :-D

This is good news.

Cheers,
Mark


-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

On Monday 04 June 2007, Daniel P. Berrange wrote:
> On Mon, Jun 04, 2007 at 01:13:01PM -0500, Nate Carlson wrote:
> > On Mon, 4 Jun 2007, Mark Williamson wrote:
> > >Not sure.  A few interesting things are currently in progress with
> > >respect to kernel versions; the Xen paravirt-ops patches may be going
> > >into the next release mainline Linux (maybe?  I think?) at which point
> > >you'll be able to build domU kernels directly from kernel.org sources.
> >
> > Just curious - can you also build dom0?  :)
>
> No. AFAIK, at this point in time, Xen paravirt_ops is DomU only.

I think paravirt_ops support for dom0 is a topic for future work...  The 
backend drivers will need to be reviewed until they pass inspection by the 
LKML guys, plus the other dom0-specific functionality.  I'm not familiar 
enough with the code to know how much work this is going to be.

AFAIK the paravirt-ops code is a fairly functionally complete domU 
implementation though; supports SMP guests etc.  Because it's using paravirt 
ops it should be possible to boot the same kernel natively, paravirtualised 
in a Xen domU and paravirtualised in VMware using VMI.

Cheers,
Mark

-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Mark Williamson wrote:
> Not sure.  A few interesting things are currently in progress with respect to 
> kernel versions; the Xen paravirt-ops patches may be going into the next 
> release mainline Linux (maybe?  I think?)
>   

This time for sure.  Well, its queued up in -mm and doesn't seem to have
caused any breakage there, and Linus has declared the patches "not too bad".

    J

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

> This time for sure.  Well, its queued up in -mm and doesn't seem to have
> caused any breakage there, and Linus has declared the patches "not too
> bad".

Awesome.  Is the plan to push some dom0 patches upstream eventually, or will 
that wait for a later date?  Hopefully the major part of the operation is 
done with, though: distros can ship with paravirt domU compatibility by 
default.

Cheers,
Mark

-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

Re: 2.6.20.3 to be the next kernel?

[Daniel P. Berrange]

On Mon, Jun 04, 2007 at 08:58:37PM +0100, Mark Williamson wrote:
> > This time for sure.  Well, its queued up in -mm and doesn't seem to have
> > caused any breakage there, and Linus has declared the patches "not too
> > bad".
> 
> Awesome.  Is the plan to push some dom0 patches upstream eventually, or will 
> that wait for a later date?  Hopefully the major part of the operation is 
> done with, though: distros can ship with paravirt domU compatibility by 
> default.

Don't forgot 64-bit support too... Current patchset is 32-bit only thus
far, so we're unable to switch to pv ops by default in Fedora even once
this current stuff is merged.  And then ppc...ia64...

Dan.
-- 
|=- Red Hat, Engineering, Emerging Technologies, Boston.  +1 978 392 2496 -=|
|=-           Perl modules: http://search.cpan.org/~danberr/              -=|
|=-               Projects: http://freshmeat.net/~danielpb/               -=|
|=-  GnuPG: 7D3B9505   F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505  -=| 

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Mark Williamson wrote:
>> This time for sure.  Well, its queued up in -mm and doesn't seem to have
>> caused any breakage there, and Linus has declared the patches "not too
>> bad".
>>     
>
> Awesome.  Is the plan to push some dom0 patches upstream eventually, or will 
> that wait for a later date?  Hopefully the major part of the operation is 
> done with, though: distros can ship with paravirt domU compatibility by 
> default.
>   

I think the vendor position is that pv-ops xen is a nice start that just
makes things more complex for them in the short term, since it only
handles one quarter of the dom0-domU/32-64bit Xen parameter space.  But
it's a start.

I'm starting on dom0 stuff now (with a minor detour via getting Xen to
boot bzImage kernels), mostly so I can get a feel for what the dom0
pv-ops patches will look like.

I'm also hoping that (ideally) x86-32 and -64 trees get merged in time
to make the 64-bit xen support trivial, but given that that's unlikely
anytime soon, I'm waiting for the 64-bit lguest/pv-ops people to get
x86-64 pv-ops in sync with 32-bit before starting on the Xen pieces.  Of
course if that takes too long (ie, longer than dom0) I might have to get
more actively involved.

    J

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Daniel P. Berrange wrote:
> Don't forgot 64-bit support too... Current patchset is 32-bit only thus
> far, so we're unable to switch to pv ops by default in Fedora even once
> this current stuff is merged.  And then ppc...ia64...

Case in point.  Patches accepted ;)

    J

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

Jeremy,

Hope you don't mind me picking your brains a bit whilst we're about it...

I was wondering how you handle PAE in pv-ops.  Presumably you just toggle 
based on what the kernel was built for?  It seems like PV-ops ought in 
principle to be able to make the kernel bimodal.  I'm guessing the 
appropriate response to my question is probably a simple "patches 
welcome" ;-)

Cheers,
Mark

-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!

Re: 2.6.20.3 to be the next kernel?

[Tim Post]

On Mon, 2007-06-04 at 19:01 +0100, Mark Williamson wrote:
> > I found this commitment from a development fedora repo :
> >
> > http://www.redhat.com/archives/fedora-extras-commits/2007-May/msg02474.html
> >
> > On may 10, looks like fedora is about to go to 2.6.20.3 for its xen
> > kernel (sometime in the not too distant future).
> 
> I believe the Fedora folks forward-port the Xen patches to newer kernels, so 
> this doesn't necessarily indicate that the mainline Xen tree will be moving 
> to this kernel version.

For Dan / Jeremy -

The commit archive interested me because I'm also active in ext3cow
which patches a 2.6.20.3 kernel. I hope a srpm or public cvs access to
that repo comes out soon :)

Rather than bisect ext3 back to 2.6.18 from 2.6.20.3 against a patched
copy of 2.6.20.3 to see what changed, I'd rather wait for this one. It
should 'just work' without introducing bugs due to back porting :) Just
thinking about doing that makes me cross eyed.

> > Where is the next (planned) hop in Linux kernels going to lead?
> 
> Not sure.  A few interesting things are currently in progress with respect to 
> kernel versions; the Xen paravirt-ops patches may be going into the next 
> release mainline Linux (maybe?  I think?) at which point you'll be able to 
> build domU kernels directly from kernel.org sources.

Cool, Virtual Legos :) And congrats on a "not bad" from Linus especially
on such a major introduction, I've seen many people running from vger
with their tail between their legs.

> Secondly we're going to move away from the sparse tree layout (yay) and 
> towards having a separate repository for the XenLinux tree.

That really makes things easier, a number of people already dismantle
the Mercurial tree on landing to separate the two.

I'm looking for changeset 15204 in the xen-unstable Mercurial tree but
it doesn't look like its been committed yet - it ends @ (tip) 15200 /
bd3d6b4c52ec , I'll look forward then to the next commits :)

Wow, Glad I asked :) Thanks Mark, Dan, Jeremy and the rest who replied,
and everyone for such a great free product.

Best,
--Tim

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Mark Williamson wrote:
> I was wondering how you handle PAE in pv-ops.  Presumably you just toggle 
> based on what the kernel was built for?  It seems like PV-ops ought in 
> principle to be able to make the kernel bimodal.  I'm guessing the 
> appropriate response to my question is probably a simple "patches 
> welcome" ;-)

Indeed.

Actually, its something I gave some thought to, but unfortunately it
isn't that simple.  The main problem is that the types of
pte_t/pmd_t/pgd_t change from 32 to 64 bit, and all the pmd folding
stuff is still done at compile-time.

I was also thinking about always using the PAE forms of the structures,
and actually do the conversion just as we read/write the entries.  But
even then, the various pagetable accessors/traversal functions know how
large the entries are and how many levels the pagetables have, etc.

The only slightly workable approach I thought of was to maintain a
sort-of in-kernel shadow pagetable scheme, which maintains parallel PAE
and non-PAE pagetables.  But that's hardly elegant, and poses all sorts
of its own problems (like propagating the hardware-set access/modified
bits properly, for example).

So, its all a bit of an open question.

    J

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Tim Post wrote:
> Cool, Virtual Legos :) And congrats on a "not bad" from Linus especially
> on such a major introduction, I've seen many people running from vger
> with their tail between their legs.
>   

Thanks, but there's a subtle semantic difference between "not bad" and
"not too bad".  The latter suggests that its actually bad, but not so
bad that he'll veto it.  "Not bad" would be too much to hope for ;)

    J

Re: 2.6.20.3 to be the next kernel?

[Jeremy Fitzhardinge]

Jeremy Fitzhardinge wrote:
> So, its all a bit of an open question.
>   

Oh, I meant to say, the UNSW/Gelato folks have been proposing a more
abstracted pagetable interface which should hide all the architectural
details of the pagetable from the core kernel.  That's probably the
right way to go to make PAE/non-PAE runtime switchable (their actual
goal is to make efficient use of the myriad ia64 pagetable modes, which
makes PAE rather trivial by comparison).  But I don't know what the
state of those patches is - I don't think they've been posted in a while.

    J

Re: 2.6.20.3 to be the next kernel?

[Mark Williamson]

> > So, its all a bit of an open question.
>
> Oh, I meant to say, the UNSW/Gelato folks have been proposing a more
> abstracted pagetable interface which should hide all the architectural
> details of the pagetable from the core kernel.  That's probably the
> right way to go to make PAE/non-PAE runtime switchable (their actual
> goal is to make efficient use of the myriad ia64 pagetable modes, which
> makes PAE rather trivial by comparison).  But I don't know what the
> state of those patches is - I don't think they've been posted in a while.

Mmmmm.  I was actually about to mention those patches too.  They seem like 
they'd be the cleanest solution to this, actually, if they were ever 
accepted.

If anyone every wanted to do something really crazy like a 32/64 dual mode 
kernel, I guess this would also be the way to go.

Cheers,
Mark

-- 
Dave: Just a question. What use is a unicyle with no seat?  And no pedals!
Mark: To answer a question with a question: What use is a skateboard?
Dave: Skateboards have wheels.
Mark: My wheel has a wheel!