Re: [Qemu-devel] [mttcg] cputlb: Use async tlb_flush_by_mmuidx

[alvise rigo]

On Mon, Feb 29, 2016 at 2:55 PM, Peter Maydell <address@hidden> wrote:
> On 29 February 2016 at 13:50, Paolo Bonzini <address@hidden> wrote:
>>
>>
>> On 29/02/2016 14:21, Peter Maydell wrote:
>>> On 29 February 2016 at 13:16, Alvise Rigo <address@hidden> wrote:
>>>> > As in the case of tlb_flush(), also tlb_flush_by_mmuidx has to query the
>>>> > TLB flush if it targets another VCPU. To accomplish this, a new async
>>>> > work has been added, together with a new TLBFlushByMMUIdxParams. A
>>>> > bitmap is used to track the MMU indexes to flush.
>>>> >
>>>> > This patch applies to the multi_tcg_v8 branch.
>>> What's the API for a target CPU emulation to say "and now I must
>>> wait for the TLB op to finish" before completing this guest
>>> instruction?
>>
>> My proposal has been for a while for DMB to put the CPU in a halted
>> state (remote TLB callbacks then can decrement a counter and signal
>> cpu_halt_cond when it's zero), but no one has implemented this.
>
> Yeah, that's the other approach -- really split the things that can
> be async and do real "wait for completion" at points which must
> synchronize. (Needs a little care since DMB is not the only such point.)
> An initial implementation that does an immediate wait-for-completion
> is probably simpler to review though, and add the real asynchrony
> later. And either way you need an API for the target to wait for
> completion.

OK, so basically being sure that the target CPU performs the flush
before executing the next TB is not enough. We need a sort of feedback
that the flush has been done before emulating the next guest
instruction. Did I get it right?

Thank you,
alvise

>
> thanks
> -- PMM