Re: [Qemu-devel] [PATCH v6 13/19] cputlb: atomically update tlb fields used by tlb_reset_dirty

[Richard Henderson]

On 11/10/2016 07:00 PM, Alex Bennée wrote:
> I should probably expand that to default to false in the case of (sizeof
> target_ulong > sizeof void *) when we don't have CONFIG_ATOMIC64.
>
> Then if the user does force mttcg on they will quickly get an assert
> although maybe we want to report that in a nicer way?

While forcing mttcg is good for testing, small hosts will definitely fail, so 
there's not point in even trying.  We should report it in a nicer way.

We shouldn't be checking sizeof(void*), but checking TCG_TARGET_REG_BITS.  That 
says how wide the host registers actually are, as opposed to the memory model 
in effect -- think x86_64 in x32 mode and the like.

If the host register size is smaller than the guest register size, we should 
force disable mttcg, regardles of CONFIG_ATOMIC64, because e.g. normal 64 bit 
loads and stores won't be atomic.


r~