[CentOS-virt] The future of blktap in Xen4CentOS

Fri May 8 10:01:39 UTC 2015
Pasi Kärkkäinen <pasik at iki.fi>


On Wed, May 06, 2015 at 03:27:54PM +0100, George Dunlap wrote:
> On Tue, May 5, 2015 at 2:07 PM, Johnny Hughes <johnny at centos.org> wrote:
> > WRT a new kernel .. the 3.18.x kernel has been selected to be maintained
> > as an LTS until Jan 2017.  I think we should start work on rebuilding
> > this kernel and using it in the x4c projects.
> >
> > https://lwn.net/Articles/636289/
> >
> > Thoughts?
> >
> > I will grab this kernel and try to roll in all the patches, etc. and see
> > if I can make it work.
> I did a little bit of work on building the 3.18 kernel a couple of
> weeks ago, but it looks like the blktap2.5 kernel code we were using
> before has totally bitrotted.
> I see three ways forward:
> 1. Try to do a "deep port" of the functionality, continuing to
> maintain it going forward.

How badly has the blktap 2.5 kernel code bitrotted? What areas are the most problematic?
Any idea how much work would it be to get it working with 3.18 LTS ? 

> 2. Start using the blktap3 code, which requires no kernel driver.
> 3. Drop blktap support, and ask people to use qdisk for vhd.
> Re #1, I'm not really excited about the idea of trying to keep porting
> those patches forward (and finding and fixing bugs in the forward-port
> each time).
> Re #2, this would require getting the blktap3 stuff working upstream
> and then backporting it.

Yeah.. likely more work than the "deep port" of blktap 2.5 kernel code..

> #3 would be by far the easiest.

But not very nice.. 
> In parallel to the Xen for C7 work I've been doing, I've been working
> on getting upstream Xen to build against XenServer's blktap code, with
> then aim of moving to blktap3.  If we want to go with #2, then a major
> kernel update will have to wait until that work is done.
>  -George


-- Pasi