On 05/02/2024 11:42, Julien Grall wrote:
>
>
> Hi Alex,
>
> On 31/01/2024 17:50, Alex Bennée wrote:
>> While debugging VirtIO on Arm we ran into a warning due to memory
>> being memcpy'd across MMIO space. While the bug was in the mappings
>> the warning was a little confusing:
>>
>> (XEN) d47v2 Rn should not be equal to Rt except for r31
>> (XEN) d47v2 unhandled Arm instruction 0x3d800000
>> (XEN) d47v2 Unable to decode instruction
>>
>> The Rn == Rt warning is only applicable to single register load/stores
>> so add some verification steps before to weed out unexpected accesses.
>>
>> I updated the Arm ARM reference to the online instruction decoding
>> table which will hopefully be more stable than the Arm ARM section
>> numbers.
NIT: commit msg should be written in imperative mood.
>
> I am not sure if the links to the Arm websites are stable. But from
> past, experience, URL tends to disappear after a while. This is why we
> went with the section + the Arm spec.
>
> This also has the advantage that we can check any differences between
> version. So my preference is to stick the Arm ARM reference. Bertrand,
> Michal, Stefano, any opinions?
I would prefer if we refer to Arm ARM instead of links to Arm websites.
Also, looking at Arm ARM J.a C4.1.88 (Loads and Stores) the encoding is a bit
different compared to that website and includes op3 and op4 (op2 is 24:23).
~Michal
