On 7 August 2015 at 13:33, Peter Maydell <peter.mayd...@linaro.org> wrote:
> Guest CPU TLB maintenance operations may be sufficiently
> specialized to only need to flush TLB entries corresponding
> to a particular MMU index. Implement cputlb functions for
> this, to avoid the inefficiency of flushing TLB entries
> which we don't need to.
>
> Signed-off-by: Peter Maydell <peter.mayd...@linaro.org>
> ---
> cputlb.c | 81
> +++++++++++++++++++++++++++++++++++++++++++++++++
> include/exec/exec-all.h | 47 ++++++++++++++++++++++++++++
> 2 files changed, 128 insertions(+)
>
> diff --git a/cputlb.c b/cputlb.c
> index a506086..a1996ba 100644
> --- a/cputlb.c
> +++ b/cputlb.c
> @@ -69,6 +69,39 @@ void tlb_flush(CPUState *cpu, int flush_global)
> tlb_flush_count++;
> }
>
> +static inline void v_tlb_flush_by_mmuidx(CPUState *cpu, va_list argp)
> +{
> + CPUArchState *env = cpu->env_ptr;
> +
> +#if defined(DEBUG_TLB)
> + printf("tlb_flush_by_mmuidx %d:\n", mmu_idx);
> +#endif
This debug tracing doesn't compile if enabled -- it was written
to go with my initial implementation which took a single
mmu_idx rather than varargs, and I forgot to update it. I'll
send out a v2 shortly.
thanks
-- PMM
