On 18 October 2017 at 23:03, Stefano Stabellini <sstabell...@kernel.org> wrote:
> WFI/E are 4 bytes long: set ARM_EL_IL_SHIFT in the syndrome.
>
> Signed-off-by: Stefano Stabellini <sstabell...@kernel.org>
>
> diff --git a/target/arm/internals.h b/target/arm/internals.h
> index 1f6efef..cf8c966 100644
> --- a/target/arm/internals.h
> +++ b/target/arm/internals.h
> @@ -398,6 +398,7 @@ static inline uint32_t syn_breakpoint(int same_el)
> static inline uint32_t syn_wfx(int cv, int cond, int ti)
> {
> return (EC_WFX_TRAP << ARM_EL_EC_SHIFT) |
> + (1 << ARM_EL_IL_SHIFT) |
> (cv << 24) | (cond << 20) | ti;
> }
Hmm. What we do now is definitely wrong, but WFI and WFE can be 2 bytes:
there is a T1 Thumb encoding that is 2 bytes.
HELPER(wfi) doesn't get that right, though:
if (target_el) {
env->pc -= 4;
raise_exception(env, EXCP_UDEF, syn_wfx(1, 0xe, 0), target_el);
}
So I think that HELPER(wfi) needs to be passed an extra
parameter is_16bit, which it can then use both in its adjustment
of env->pc and to pass as an extra parameter to syn_wfx(),
which is then syn_wfx(int cv, int cond, int ti, bool is_16bit).
(In theory HELPER(wfe) should also be passed is_16bit, but
since it doesn't currently ever raise an exception it
doesn't matter.)
thanks
-- PMM
