CAS implementation may be broken

[Toby Douglass]

Jamie Lokier wrote:
> Toby Douglass wrote:

>> Interestingly, I saw a day or two ago that there is a double-word 
>> version of LDREX.  The atomic-ptr project, which forms the basis of the 
>> GCC garbage collector, relies on this; it does not in fact use LL/SC on 
>> ARM, but rather uses double-word CAS using a pointer-counter pair.

> Note that pointer-counter is not really reliable.

[snip case]

> Of course even if the counter wraps, you still have to be 1/2^32
> unlucky to see the same value.  But that's enough to make it unreliable.

I concur, but it's not a problem on 64-bit.