Re: Broken ARM (and powerpc ?) futex wrt memory barriers

[Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>]

* Russell King - ARM Linux (linux@arm.linux.org.uk) wrote:
> On Mon, May 25, 2009 at 05:45:33PM -0400, Mathieu Desnoyers wrote:
> > I think we also have to deal with futexes. See
> > arch/arm/include/asm/futex.h :
> > 
> > 1 - 
> > 
> > #ifdef CONFIG_SMP
> > 
> > #include <asm-generic/futex.h>
> > 
> > #else /* !SMP, we can work around lack of atomic ops by disabling
> > preemption */
> > 
> > (arm-specific code here, seems to deal with futexes)
> > 
> > #endif
> > 
> > -> is it just me or this ifdef condition is the exact opposite of what
> > it should be ? I thought those generic futexes were for UP-only
> > systems...
> 
> No, it is correct.  Look deeper - the generic stuff implements a whole
> load of futex stuff as returning -ENOSYS - it's not implemented.
> 
> What you see in ARM is the UP code implementing the set/add/or/andn/xor
> cmpxchg operations.  However, this code is most certainly not SMP safe
> (and I couldn't persuade the folk writing it to come up with a SMP
> version.)  So, on ARM futex is supported on UP systems, but not SMP
> systems until someone gets around to fixing this.

Ah, yes, you are right. I did not realize ARM SMP support was at such an
early stage. I think there is still some work ahead before those SMP
boards can be fully supported, but I'm glad we're making progress. :)

Mathieu


-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68