[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: (OT) Re: 2.6.20 for FC6?



Daniel P. Berrange wrote:
On Wed, Feb 07, 2007 at 09:39:51AM +0100, dragoran wrote:
Dave Jones wrote:
On Tue, Feb 06, 2007 at 03:08:04PM -0600, Gilbert Sebenste wrote:
On Tue, 6 Feb 2007, Dave Jones wrote:

It gets better...
http://www.devxnews.com/article.php/3658001

"Pratt also explained that Xen is no longer actively seeking
inclusion in the mainline Linux kernel either."
Oh, swell. So a 2.6.20 kernel for FC5 is DOA. :-(
Not really. The rebases over the last year or two have been done by
Red Hat, not Xensource.  By the time the 'official' Xen tree is
updated, it's based on some six month old kernel which is uninteresting.
So sit tight and wait for the Xen team to do their magic.
(Largely the work of Juan Quintela, who succeeds in doing in a month
what takes Xensource half a year to do)

I don't know how Juan handles this but wouldn't it be easier to port xen during a kernel development cycle and not after the kernel is released? If there are some last minute changes they wont be that big. I assume that this will save a lot of time (if not already done).

Juan handles a hell of alot of work - its not merely tracking upstream
LKML, but also tracking upstream xen-devel. Doing this for i386, x86_64
and ia64, and many of the really nasty merge issues are low level hardware
stuff. Add to that Juan's maintaining upto 6 kernel trees - xen 3.0.3 against 2.6.18, .19, .20, likewise for xen 3.0.4 - providing updates for 3 Fedora releases FC5, FC6 and rawhide. At the same time we are continually
pushing upstream xen-devel to get onto recent kernels to make this work
easier, as well as having more folks working on paravirt_ops and a Xen
paravirt_ops impl as a 2nd strategy for getting a mainline Xen tree.

So yes, having Xen out of tree is incredibly painful for everyone concerned
and we are pursuing multiple angles to reduce this pain and hopefully get
to a state where Xen is in sync with LKML, and preferrably merged. Its a
large body of code so its not a quick or easy process :-(

Regards,
Dan.
ok thanks for that info


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]