Re: [PATCH] cputopology: Always define CPU topology information [4th try]

[Ben Hutchings <bhutchings@solarflare.com>]

Nathan Lynch wrote:
> Hi Ben-
> 
> Ben Hutchings wrote:
> > Not all architectures and configurations define CPU topology information.
> > This can result in an empty topology directory in sysfs, and requires
> > in-kernel users to protect all uses with #ifdef - see
> > <http://marc.info/?l=linux-netdev&m=120639033904472&w=2>.
> > 
> > The documentation of CPU topology specifies what the defaults should be
> > if only partial information is available from the hardware.  So we can
> > provide these defaults as a fallback.
> 
> I've been looking at adding topology information to powerpc and I came
> across this.
> 
> I understand the need for fallback definitions of the topology APIs
> within the kernel, but I'm not sure I agree with exposing these things
> in sysfs unconditionally -- the default values for physical_package_id
> and core_id don't really make sense on powerpc (and other non-x86
> architectures, I suspect).

In what way are they wrong?

> Would you object to a patch which exposes in sysfs only the topology
> information which the architecture provides?

I was primarily concerned with having the fallbacks available in-kernel.
However, I don't think you will be doing user-space any favours by
requiring checks for missing attributes for ever.

Ben.

-- 
Ben Hutchings, Senior Software Engineer, Solarflare Communications
Not speaking for my employer; that's the marketing department's job.