Re: Xen, Squeeze, and Beyond
On Sun, 2010-02-28 at 01:03 +0300, William Pitcock wrote:
> ----- "Josip Rodin" <joy@entuzijast.net> wrote:
>
> > On Sat, Feb 27, 2010 at 01:23:07AM +0300, William Pitcock wrote:
> > > I am looking into packaging xenner already as a backup plan if I
> > cannot
> > > manage to fix some major reentrancy problems in the Xen dom0 code
> > > (Xensource 2.6.18 patches, the pvops stuff has it's own share of
> > problems
> > > and needs more evaluation).
> >
> > The .18 dom0 patches are well on their way out from the perspective
> > of
> > both Debian and Xen upstream, so you might want to shift focus to the
> > pvops
> > branch instead.
>
> I am well aware of that. However, the pvops branch has several critical
> bugs:
>
> - On a 8-way system, it reports 259GHz CPUs for all cores when booted under
> Xen;
> - The paravirtualized clock is 4 times slower then it should be in dom0
> mode;
> - The same reentrancy issues exist, as
It is very much worth taking these issues to the xen-devel mailing list.
> the pvops work is mostly Jeremy forward porting the 2.6.18 code
That's not completely true. Things like the backend drivers are largely
forward ports (although new development work is also occurring in this
tree these days) but the core infrastructure (memory management,
interrupts etc) are rewritten from scratch for the pvops kernels.
> There are also no pvops dom0 kernel packages shipped by Debian yet, at
> least through official channels.
>
> While you are correct that pvops is the future, right now it's no better
> reliability-wise then the 2.6.18 xensource patches... unfortunately tracking
> these reentrancy bugs (mostly deadlocks) down is a massive pain in the ass.
pvops may be "no better" than the 2.6.18 patches but pvops is the one
where there will be the possibility of gaining momentum/interest to get
the issues fixed.
Ian.
--
Ian Campbell
Current Noise: Rammstein - Spring
It's not reality that's important, but how you perceive things.
Reply to: