Re: [Qemu-devel] [PATCH 0/2] Implement YIELD to yield in ARM and Thumb translators

[Peter Crosthwaite]

On Fri, Jun 26, 2015 at 7:28 AM, Peter Maydell <address@hidden> wrote:
> On 26 June 2015 at 15:25, Paolo Bonzini <address@hidden> wrote:
>>
>>
>> On 26/06/2015 16:04, Peter Maydell wrote:
>>> Ping?
>>
>> Oh, YIELD was what I was thinking of in the thread about IPIs, not WFE
>> (which would require SEV on the other core)...
>
> Linux doesn't in practice put YIELD insns into its busy loops,
> so this series is more of a completeness thing than an actual
> attempt to solve the problem in the other thread.
>

So it does for AA64 now:

https://lkml.org/lkml/2015/3/2/672

Should we submit the same patch to Linux for AA32?

> I think the right fix for the other thread's problems is that
> we should not allow any TCG CPU to run for an unbounded period
> of time -- we should always switch to another CPU after some
> period of time if the CPU hasn't ended up coming back out to
> the top-level loop of its own accord. That's definitely post-2.4
> material though.

Agree. The guest should not be able to break the emulation.

Regards,
Peter

>
> -- PMM
>