CONFIG_SMP or !CONFIG_SMP... that is the question. WAS: something about ia64 that nobody would read :-)

CONFIG_SMP or !CONFIG_SMP... that is the question. WAS: something about ia64 that nobody would read :-)

[Magenheimer, Dan (HP Labs Fort Collins)]

> From: Ian Pratt [mailto:m+Ian.Pratt@cl.cam.ac.uk] 

> It's daft that we even have a CONFIG_SMP option in Xen. It spends most
> of its time broken because no-one using x86 builds it, and you won't
> even be able to buy any non SMP hardware before long...
> 
> I'd vote for expunging CONFIG_SMP.

I've heard its not uncommon when debugging nasty problems on Linux
to turn off SMP as it simplifies the world considerably.

I'm not sure this same argument applies to Xen, but there's enough
kernel hackers on this list that I thought it would be interesting
to open this up for discussion.

Granted, turning off SMP on Xen/x86 doesn't even compile right
now, but that should be fixable.

So... opinions anyone?  Is keeping CONFIG_SMP (potentially) useful
or not?

Re: CONFIG_SMP or !CONFIG_SMP... that is the question. WAS: something about ia64 that nobody would read :-)

[Christian Limpach]

On 5/11/05, Magenheimer, Dan (HP Labs Fort Collins)
<dan.magenheimer@hp.com> wrote:
> > From: Ian Pratt [mailto:m+Ian.Pratt@cl.cam.ac.uk]
> 
> > It's daft that we even have a CONFIG_SMP option in Xen. It spends most
> > of its time broken because no-one using x86 builds it, and you won't
> > even be able to buy any non SMP hardware before long...
> >
> > I'd vote for expunging CONFIG_SMP.
> 
> I've heard its not uncommon when debugging nasty problems on Linux
> to turn off SMP as it simplifies the world considerably.

I would either run on non-SMP hardware, use the nosmp option or if
that's gone, just nobble the code that discovers multiple CPUs -- all
these are definitely preferable to compiling Xen without CONFIG_SMP
since that might change the code so much that the bug might not occur
anymore.

    christian

Re: CONFIG_SMP or !CONFIG_SMP... that is the question. WAS: something about ia64 that nobody would read :-)

[Andrei Petrov]

On Wed, May 11, 2005 at 12:46:08PM -0700, Magenheimer, Dan (HP Labs Fort Collins) wrote:
> 
> Granted, turning off SMP on Xen/x86 doesn't even compile right
> now, but that should be fixable.
> 
> So... opinions anyone?  Is keeping CONFIG_SMP (potentially) useful
> or not?

Depends on target application area, if xen will be used in appliances or
embedded systems, or laptops (think QOS, securi1y) then keeping CONFIG_SMP
(!CONFIG_SMP even) would be an advantage for the project.

--Andrei

RE: CONFIG_SMP or !CONFIG_SMP... that is the question. WAS:something about ia64 that nobody would read :-)

[Ian Pratt]

 
 
> > I'd vote for expunging CONFIG_SMP.
> 
> I've heard its not uncommon when debugging nasty problems on 
> Linux to turn off SMP as it simplifies the world considerably.

Seting maxcpus=1 has basically the same dynamic effect. The only point
of having the build option is to reduce the number of LOCK'ed
operations.

Ian
 
> I'm not sure this same argument applies to Xen, but there's 
> enough kernel hackers on this list that I thought it would be 
> interesting to open this up for discussion.
> 
> Granted, turning off SMP on Xen/x86 doesn't even compile 
> right now, but that should be fixable.
> 
> So... opinions anyone?  Is keeping CONFIG_SMP (potentially) 
> useful or not?
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
>