Re: [PATCH] risu: don't do a full register compare for OP_SIGILL

[Richard Henderson]

On 6/23/20 9:17 AM, Alex Bennée wrote:
> 
> Peter Maydell <peter.maydell@linaro.org> writes:
> 
>> On Tue, 23 Jun 2020 at 16:07, Alex Bennée <alex.bennee@linaro.org> wrote:
>>>
>>> OP_SIGILL means we have an unexpected invalid operation. If this is a
>>> load or store the register state may be un-rectified pointing at the
>>> memblock so would be invalid. In this case just compare the PC and
>>> make sure the other end also faulted at the same place.
>>
>> In case of mismatch of the PC do we still print the full register
>> dump?

No.  If we want that, we should do something else, like remember that the
memory pointer is in use and zap it out before reporting the register set.

But, generally, if we see SIGILL, then we have not actually executed anything,
so the register state doesn't matter too much.  What probably does want
reporting in this case is the insn opcode.


r~