Re: [PATCH] WorkStruct: Implement generic UP cmpxchg() where an archdoesn't support it

[Linus Torvalds]

On Fri, 8 Dec 2006, Russell King wrote:
>
> Yes you can.  Well, you can on ARM at least.  Between the load exclusive
> you can do anything you like until you hit the store exclusive.  If you
> haven't touched the location (with anything other than another load
> exclusive) and no other CPU has issued a load exclusive, your store
> exclusive succeeds.

Is that actually true?

Almost all LL/SC implementations have granularity rules, where "touch the 
location" is not a byte-granular thing, but actually ends up being 
something like "touch the same cachline".

They also often have _other_ rules like: "the cacheline has to stay in the 
L1 in exclusive state" etc. Which means that in a direct-mapped L1 cache, 
you can't even load anything that might be in the same way, because it 
would cause a cache eviction that invalidates the SC.

It's possible that ARM has really strong LL/SC, but quite frankly, that 
sounds unlikely. I've never heard of anybody ever _architecturally_ saying 
that they support that strong requirements, even if certain micro- 
architectures might actually support stronger semantics than the ones 
guaranteed by the architectural rules.

			Linus
-
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/