[8u] RFR: nativeInstruction atomicity and correctness of patching

[Andrew Haley]

On 03/30/2016 04:15 PM, Anton Kozlov wrote:
> ICache::invalidate_word used here not to provide coherency between
> Icache and Dcache, but to maintain in-order visibillity of code
> modification.

What exactly are you trying to synchronize?  Any code modifications
must have their own icache invalidation, which does a full DSB.  But
this call does not modify any code.

> I mean, if NativeTrampolineCall::set_destination_mt_safe
> called before another code modification, effect of it will be observed
> by another cpu before effect of 2nd code modification.

Only if that other CPU executes a read fence.

> This is achieved by implementation of ICache::invalidate_word, which
> is just call to __clear_cache syscall wrapper.

I don't think there's any point.  It's not instructions but data, so
ICache::invalidate_word is inappropriate.

Andrew.