Re: [PATCH v2 1/3] arm: handle demuxed ID registers

[Cornelia Huck <cohuck@redhat.com>]

On Wed, Apr 01 2026, Eric Auger <eric.auger@redhat.com> wrote:

> On 4/1/26 10:17 AM, Cornelia Huck wrote:
>> On Thu, Feb 05 2026, Richard Henderson <richard.henderson@linaro.org> wrote:
>>> I'm not keen on this.  Surely we don't expect all multiplexed ID registers to have the 
>>> same number of sub-registers?  A rectangular matrix isn't a great representation.  Also, 
>>> you're wasting a hole in idregs[] with CSSIDR_EL1_IDX.
>>>
>>> Why not just have the N sub-registers occupy N entries within the regular idregs[]? 
>>> Something like
>>>
>>> #define DEF_MUX(NAME, OP0, OP1, CRN, CRM, OP2, NUM) \
>>>      NAME##_DEMUX_IDX,                               \
>>>      NAME##_DEMUX_LAST = NAME##_DEMUX_IDX + NUM - 1,
>>>
>>> #define SET_IDREG_DEMUX(ISAR, REG, INDEX, VALUE)                        \
>>>      ({                                                                  \
>>>          ARMISARegisters *i_ = (ISAR);                                   \
>>>          i_->idregs_demux[REG ## _EL1_DEMUX_IDX + INDEX] = VALUE;        \
> I guess we would use regular i_-> idregs[] here, right? 

Yes.

>
> Thanks
>
> Eric
>>>      })
>>>
>>> etc.
>> Yeah, I was not overly excited about my solution, either. I'll go and
>> play with this, thanks.
>>
>>