Re: [RFC 00/15] x86_64: Optimize percpu accesses

[Jeremy Fitzhardinge]

Eric W. Biederman wrote:
> Jeremy Fitzhardinge <jeremy@xxxxxxxx> writes:
>
>   
> > Eric W. Biederman wrote:
> >     
> > > I think we can get away with just simply realloc'ing the percpu area
> > > on each cpu.  No fancy table manipulations required.  Just update
> > > the base pointer in %gs and in someplace global.
> > >
> > >       
> > It's perfectly legitimate to take the address of a percpu variable and store it
> > somewhere.  We can't move them around.
> >     
>
> Really?  I guess there are cases where that makes sense.  It is a pretty
> rare case though.  Especially when you are not talking about doing it temporarily
> with preemption disabled.  There are few enough users of the API I think we can
> certainly explore the cost of forbidding in the general case of storing the
> address of a percpu variable.
>   

No, that sounds like a bad idea.  For one, how would you enforce it?  
How would you check for it?  It's one of those things that would mostly 
work and then fail very rarely.

Secondly, I depend on it.  I register a percpu structure with Xen to 
share per-vcpu specific information (interrupt mask, time info, runstate 
stats, etc).

   J
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/