Re: [PATCH] target/arm: fix IL bit for data abort exceptions

[Richard Henderson]

On 12/19/19 2:43 AM, Peter Maydell wrote:
> I think this is a regression introduced in commit 46beb58efbb8a2a32
> when we converted the Thumb decoder over to decodetree. Before that
> 16 bit Thumb insns were in a different place in the old decoder and
> the 16-bit Thumb path passed ISSIs16Bit in with its issflags.
> (We should cc: address@hidden on the fix for this.)

Oops, yes, that would be my fault.

>>>      /* ISS not valid if writeback */
>>>      if (p && !w) {
>>> -        ret = rd;
>>> +        ret = rd | (s->is_16bit ? ISSIs16Bit : 0);
>>>      } else {
>>>          ret = ISSInvalid;
>>>      }
> 
> Rather than setting an is_16bit flag, we could just use
> "dc->base.pc_next - dc->pc_curr == 2", couldn't we?

We could, yes.


r~