On Tue, 13 Oct 2020 at 18:30, Richard Henderson <richard.hender...@linaro.org> wrote: > Well, the only further comment is that, in the followup, only WLS gains the IT > block check. While I understand that's required to avoid an abort in QEMU for > this case, all three of the insns have that case as CONSTRAINED UNPREDICTABLE. > It might be worthwhile checking for IT in all of them, just to continue our > normal "unpredictable raises sigill, when easy" choice.
Maybe, but there are a lot of instructions that are unpredictable-in-an-IT-block (CPSID, CRC32B, HVC...) and our general approach seems to have been "don't check unless it would cause an actual problem". The only place I can find where we do check for this case is in trans_B_cond_thumb(), which we do for the same reason as here. thanks -- PMM
